Upload the PEFT-MLLM models

CheckGuard Models/wholeimage/amount/finetune_lora_llava_mistral.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+
+deepspeed llava/train/train_mem.py \
+    --lora_enable True --lora_r 128 --lora_alpha 256 --mm_projector_lr 2e-5 \
+    --deepspeed ./scripts/zero3.json \
+    --model_name_or_path liuhaotian/llava-v1.6-mistral-7b \
+    --version mistral_instruct \
+    --data_path /home/larry5/project/LLaVA-1.6-ft/data/peft/amount/modified_path_to_train_val_human-gpt-whole-check.json \
+    --image_folder /home/larry5/project/LLaVA-1.6-ft/data/data/ \
+    --vision_tower openai/clip-vit-large-patch14-336 \
+    --mm_projector_type mlp2x_gelu \
+    --mm_vision_select_layer -2 \
+    --mm_use_im_start_end False \
+    --mm_use_im_patch_token False \
+    --mm_patch_merge_type spatial_unpad \
+    --image_aspect_ratio anyres \
+    --group_by_modality_length False \
+    --bf16 False \
+    --fp16 True \
+    --output_dir /home/larry5/project/LLaVA-1.6-ft/scripts_peft/mistral/lora/llava-lora-mistral-r128a256/wholeimage/amount/llava-lora-mistral-r128a256-10BS-model \
+    --num_train_epochs 1 \
+    --per_device_train_batch_size 10 \
+    --per_device_eval_batch_size 1 \
+    --gradient_accumulation_steps 1 \
+    --evaluation_strategy "no" \
+    --save_strategy "steps" \
+    --save_steps 500 \
+    --save_total_limit 5 \
+    --learning_rate 2e-5 \
+    --weight_decay 0. \
+    --warmup_ratio 0.05 \
+    --lr_scheduler_type "cosine" \
+    --logging_steps 1 \
+    --tf32 True \
+    --model_max_length 4096 \
+    --gradient_checkpointing True \
+    --dataloader_num_workers 4 \
+    --lazy_preprocess True \
+    --report_to wandb \

CheckGuard Models/wholeimage/amount/llava-lora-mistral-r128a256-10BS-model/README.md

@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: liuhaotian/llava-v1.6-mistral-7b
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.10.0

CheckGuard Models/wholeimage/amount/llava-lora-mistral-r128a256-10BS-model/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "liuhaotian/llava-v1.6-mistral-7b",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "v_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

CheckGuard Models/wholeimage/amount/llava-lora-mistral-r128a256-10BS-model/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf2247d527585fc799edb906388c1be818fe3bb61e79cbe1b59d3311b2b6e5e9
+size 708924928

CheckGuard Models/wholeimage/amount/llava-lora-mistral-r128a256-10BS-model/config.json

@@ -0,0 +1,70 @@
+{
+  "_name_or_path": "liuhaotian/llava-v1.6-mistral-7b",
+  "architectures": [
+    "LlavaMistralForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "freeze_mm_mlp_adapter": false,
+  "freeze_mm_vision_resampler": false,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "image_aspect_ratio": "anyres",
+  "image_crop_resolution": 224,
+  "image_grid_pinpoints": [
+    [
+      336,
+      672
+    ],
+    [
+      672,
+      336
+    ],
+    [
+      672,
+      672
+    ],
+    [
+      1008,
+      336
+    ],
+    [
+      336,
+      1008
+    ]
+  ],
+  "image_split_resolution": 224,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 32768,
+  "mm_hidden_size": 1024,
+  "mm_patch_merge_type": "spatial_unpad",
+  "mm_projector_lr": 2e-05,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "openai/clip-vit-large-patch14-336",
+  "mm_vision_tower_lr": 2e-06,
+  "model_type": "llava_mistral",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "rms_norm_eps": 1e-05,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "tokenizer_model_max_length": 4096,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.37.2",
+  "tune_mm_mlp_adapter": false,
+  "tune_mm_vision_resampler": false,
+  "unfreeze_mm_vision_tower": true,
+  "use_cache": true,
+  "use_mm_proj": true,
+  "vocab_size": 32000
+}

CheckGuard Models/wholeimage/amount/llava-lora-mistral-r128a256-10BS-model/non_lora_trainables.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:122abc8f0749c93d63088c8fbc3c18949d0e6fe8a9c9bc719442920c7224b9fc
+size 41961648

CheckGuard Models/wholeimage/amount/llava-lora-mistral-r128a256-10BS-model/trainer_state.json

@@ -0,0 +1,2010 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 330,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1764705882352942e-06,
+      "loss": 0.2385,
+      "step": 1
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3529411764705885e-06,
+      "loss": 0.2477,
+      "step": 2
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3529411764705885e-06,
+      "loss": 0.297,
+      "step": 3
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.529411764705883e-06,
+      "loss": 0.1694,
+      "step": 4
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.705882352941177e-06,
+      "loss": 0.1294,
+      "step": 5
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.882352941176471e-06,
+      "loss": 0.1461,
+      "step": 6
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.058823529411766e-06,
+      "loss": 0.1272,
+      "step": 7
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.058823529411766e-06,
+      "loss": 0.1176,
+      "step": 8
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.23529411764706e-06,
+      "loss": 0.0666,
+      "step": 9
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.411764705882354e-06,
+      "loss": 0.1199,
+      "step": 10
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0588235294117648e-05,
+      "loss": 0.1216,
+      "step": 11
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1764705882352942e-05,
+      "loss": 0.1258,
+      "step": 12
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2941176470588238e-05,
+      "loss": 0.0381,
+      "step": 13
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4117647058823532e-05,
+      "loss": 0.0318,
+      "step": 14
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5294117647058822e-05,
+      "loss": 0.1278,
+      "step": 15
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.647058823529412e-05,
+      "loss": 0.1014,
+      "step": 16
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7647058823529414e-05,
+      "loss": 0.0555,
+      "step": 17
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8823529411764708e-05,
+      "loss": 0.0576,
+      "step": 18
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2e-05,
+      "loss": 0.0987,
+      "step": 19
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9999496293646753e-05,
+      "loss": 0.1177,
+      "step": 20
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999798522533102e-05,
+      "loss": 0.1623,
+      "step": 21
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9995466947279753e-05,
+      "loss": 0.0899,
+      "step": 22
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9991941713187477e-05,
+      "loss": 0.0615,
+      "step": 23
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9987409878190752e-05,
+      "loss": 0.068,
+      "step": 24
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.99818718988324e-05,
+      "loss": 0.0909,
+      "step": 25
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9975328333015497e-05,
+      "loss": 0.0658,
+      "step": 26
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967779839947172e-05,
+      "loss": 0.0251,
+      "step": 27
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9959227180072216e-05,
+      "loss": 0.0526,
+      "step": 28
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9949671214996448e-05,
+      "loss": 0.0495,
+      "step": 29
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.993911290739993e-05,
+      "loss": 0.034,
+      "step": 30
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.992755332093999e-05,
+      "loss": 0.0678,
+      "step": 31
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9914993620144055e-05,
+      "loss": 0.063,
+      "step": 32
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.990143507029234e-05,
+      "loss": 0.0237,
+      "step": 33
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9886879037290385e-05,
+      "loss": 0.0773,
+      "step": 34
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9871326987531453e-05,
+      "loss": 0.0357,
+      "step": 35
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.98547804877488e-05,
+      "loss": 0.1064,
+      "step": 36
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.983724120485783e-05,
+      "loss": 0.083,
+      "step": 37
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9818710905788195e-05,
+      "loss": 0.1053,
+      "step": 38
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9799191457305767e-05,
+      "loss": 0.0343,
+      "step": 39
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.977868482582459e-05,
+      "loss": 0.1018,
+      "step": 40
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9757193077208776e-05,
+      "loss": 0.1488,
+      "step": 41
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9734718376564386e-05,
+      "loss": 0.0511,
+      "step": 42
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9711262988021322e-05,
+      "loss": 0.0643,
+      "step": 43
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.968682927450523e-05,
+      "loss": 0.0184,
+      "step": 44
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9661419697499455e-05,
+      "loss": 0.0483,
+      "step": 45
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9635036816797072e-05,
+      "loss": 0.1308,
+      "step": 46
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.960768329024301e-05,
+      "loss": 0.0618,
+      "step": 47
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.957936187346628e-05,
+      "loss": 0.0513,
+      "step": 48
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.955007541960241e-05,
+      "loss": 0.0517,
+      "step": 49
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9519826879005964e-05,
+      "loss": 0.0638,
+      "step": 50
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.948861929895336e-05,
+      "loss": 0.0841,
+      "step": 51
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.945645582333587e-05,
+      "loss": 0.0994,
+      "step": 52
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9423339692342885e-05,
+      "loss": 0.0906,
+      "step": 53
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9389274242135528e-05,
+      "loss": 0.2008,
+      "step": 54
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9354262904510544e-05,
+      "loss": 0.0152,
+      "step": 55
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9318309206554567e-05,
+      "loss": 0.0232,
+      "step": 56
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9281416770288806e-05,
+      "loss": 0.057,
+      "step": 57
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.924358931230418e-05,
+      "loss": 0.1069,
+      "step": 58
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.920483064338687e-05,
+      "loss": 0.034,
+      "step": 59
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9165144668134426e-05,
+      "loss": 0.052,
+      "step": 60
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9124535384562423e-05,
+      "loss": 0.1445,
+      "step": 61
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9083006883701688e-05,
+      "loss": 0.0578,
+      "step": 62
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.904056334918617e-05,
+      "loss": 0.0426,
+      "step": 63
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8997209056831462e-05,
+      "loss": 0.0214,
+      "step": 64
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8952948374204066e-05,
+      "loss": 0.084,
+      "step": 65
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8907785760181392e-05,
+      "loss": 0.1055,
+      "step": 66
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8861725764502557e-05,
+      "loss": 0.0333,
+      "step": 67
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.881477302731006e-05,
+      "loss": 0.0334,
+      "step": 68
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.87669322786823e-05,
+      "loss": 0.0513,
+      "step": 69
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8718208338157082e-05,
+      "loss": 0.0324,
+      "step": 70
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.866860611424609e-05,
+      "loss": 0.033,
+      "step": 71
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8618130603940386e-05,
+      "loss": 0.0379,
+      "step": 72
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856678689220701e-05,
+      "loss": 0.0403,
+      "step": 73
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851458015147673e-05,
+      "loss": 0.0489,
+      "step": 74
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846151564112294e-05,
+      "loss": 0.0599,
+      "step": 75
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.840759870693184e-05,
+      "loss": 0.0654,
+      "step": 76
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8352834780563888e-05,
+      "loss": 0.0526,
+      "step": 77
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8297229379006614e-05,
+      "loss": 0.0105,
+      "step": 78
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8240788104018824e-05,
+      "loss": 0.0394,
+      "step": 79
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8183516641566278e-05,
+      "loss": 0.0573,
+      "step": 80
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8125420761248878e-05,
+      "loss": 0.0478,
+      "step": 81
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.806650631571943e-05,
+      "loss": 0.0633,
+      "step": 82
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8006779240094024e-05,
+      "loss": 0.0423,
+      "step": 83
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7946245551354156e-05,
+      "loss": 0.0618,
+      "step": 84
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7884911347740556e-05,
+      "loss": 0.0658,
+      "step": 85
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.782278280813882e-05,
+      "loss": 0.0485,
+      "step": 86
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.775986619145697e-05,
+      "loss": 0.0468,
+      "step": 87
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7696167835994927e-05,
+      "loss": 0.0558,
+      "step": 88
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7631694158805945e-05,
+      "loss": 0.0518,
+      "step": 89
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7566451655050197e-05,
+      "loss": 0.0684,
+      "step": 90
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7500446897340408e-05,
+      "loss": 0.0304,
+      "step": 91
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7433686535079736e-05,
+      "loss": 0.0397,
+      "step": 92
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.736617729379191e-05,
+      "loss": 0.1084,
+      "step": 93
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7297925974443675e-05,
+      "loss": 0.0826,
+      "step": 94
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7228939452759666e-05,
+      "loss": 0.0309,
+      "step": 95
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7159224678529734e-05,
+      "loss": 0.0348,
+      "step": 96
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7088788674908817e-05,
+      "loss": 0.097,
+      "step": 97
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7017638537709426e-05,
+      "loss": 0.0897,
+      "step": 98
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6945781434686783e-05,
+      "loss": 0.0614,
+      "step": 99
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6873224604816753e-05,
+      "loss": 0.08,
+      "step": 100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.679997535756657e-05,
+      "loss": 0.0126,
+      "step": 101
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.672604107215848e-05,
+      "loss": 0.0644,
+      "step": 102
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6651429196826337e-05,
+      "loss": 0.0702,
+      "step": 103
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6576147248065268e-05,
+      "loss": 0.0809,
+      "step": 104
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6500202809874446e-05,
+      "loss": 0.0354,
+      "step": 105
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6423603532993074e-05,
+      "loss": 0.1429,
+      "step": 106
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.634635713412964e-05,
+      "loss": 0.1142,
+      "step": 107
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.626847139518452e-05,
+      "loss": 0.034,
+      "step": 108
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.618995416246601e-05,
+      "loss": 0.0818,
+      "step": 109
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6110813345899914e-05,
+      "loss": 0.0594,
+      "step": 110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.6031056918232642e-05,
+      "loss": 0.0958,
+      "step": 111
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.595069291422807e-05,
+      "loss": 0.0418,
+      "step": 112
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.586972942985807e-05,
+      "loss": 0.0315,
+      "step": 113
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5788174621486936e-05,
+      "loss": 0.0435,
+      "step": 114
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.570603670504969e-05,
+      "loss": 0.0596,
+      "step": 115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.570603670504969e-05,
+      "loss": 0.048,
+      "step": 116
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5623323955224404e-05,
+      "loss": 0.0352,
+      "step": 117
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5540044704598588e-05,
+      "loss": 0.0264,
+      "step": 118
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5456207342829777e-05,
+      "loss": 0.0378,
+      "step": 119
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5371820315800316e-05,
+      "loss": 0.0519,
+      "step": 120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5286892124766546e-05,
+      "loss": 0.0559,
+      "step": 121
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5201431325502332e-05,
+      "loss": 0.0708,
+      "step": 122
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5115446527437193e-05,
+      "loss": 0.0823,
+      "step": 123
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.5028946392788934e-05,
+      "loss": 0.0345,
+      "step": 124
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4941939635691036e-05,
+      "loss": 0.111,
+      "step": 125
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4854435021314766e-05,
+      "loss": 0.0284,
+      "step": 126
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4766441364986162e-05,
+      "loss": 0.1226,
+      "step": 127
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.467796753129797e-05,
+      "loss": 0.022,
+      "step": 128
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4589022433216616e-05,
+      "loss": 0.0565,
+      "step": 129
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4499615031184297e-05,
+      "loss": 0.0875,
+      "step": 130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4409754332216303e-05,
+      "loss": 0.0457,
+      "step": 131
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.431944938899363e-05,
+      "loss": 0.0776,
+      "step": 132
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4228709298950998e-05,
+      "loss": 0.0397,
+      "step": 133
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.4137543203360382e-05,
+      "loss": 0.0278,
+      "step": 134
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.4045960286410093e-05,
+      "loss": 0.0204,
+      "step": 135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.395396977427955e-05,
+      "loss": 0.0531,
+      "step": 136
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3861580934209832e-05,
+      "loss": 0.0334,
+      "step": 137
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.376880307357009e-05,
+      "loss": 0.0389,
+      "step": 138
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3675645538919884e-05,
+      "loss": 0.0571,
+      "step": 139
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3582117715067628e-05,
+      "loss": 0.0352,
+      "step": 140
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3488229024125142e-05,
+      "loss": 0.0062,
+      "step": 141
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3393988924558445e-05,
+      "loss": 0.0489,
+      "step": 142
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3299406910234917e-05,
+      "loss": 0.068,
+      "step": 143
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.3204492509466862e-05,
+      "loss": 0.0478,
+      "step": 144
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.3109255284051615e-05,
+      "loss": 0.1145,
+      "step": 145
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.3013704828308276e-05,
+      "loss": 0.0253,
+      "step": 146
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2917850768111171e-05,
+      "loss": 0.0138,
+      "step": 147
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.282170275992012e-05,
+      "loss": 0.1083,
+      "step": 148
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2725270489807637e-05,
+      "loss": 0.0708,
+      "step": 149
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2628563672483147e-05,
+      "loss": 0.0144,
+      "step": 150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2531592050314308e-05,
+      "loss": 0.0628,
+      "step": 151
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2434365392345553e-05,
+      "loss": 0.0475,
+      "step": 152
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2336893493313946e-05,
+      "loss": 0.0161,
+      "step": 153
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.223918617266245e-05,
+      "loss": 0.084,
+      "step": 154
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.2141253273550698e-05,
+      "loss": 0.0494,
+      "step": 155
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.2043104661863386e-05,
+      "loss": 0.0293,
+      "step": 156
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1944750225216363e-05,
+      "loss": 0.0837,
+      "step": 157
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1846199871960557e-05,
+      "loss": 0.0479,
+      "step": 158
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1747463530183781e-05,
+      "loss": 0.0752,
+      "step": 159
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1648551146710557e-05,
+      "loss": 0.0242,
+      "step": 160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1549472686100079e-05,
+      "loss": 0.0322,
+      "step": 161
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.145023812964237e-05,
+      "loss": 0.0812,
+      "step": 162
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1350857474352734e-05,
+      "loss": 0.0133,
+      "step": 163
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1251340731964664e-05,
+      "loss": 0.093,
+      "step": 164
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1151697927921242e-05,
+      "loss": 0.0377,
+      "step": 165
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1051939100365154e-05,
+      "loss": 0.0561,
+      "step": 166
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0952074299127451e-05,
+      "loss": 0.0556,
+      "step": 167
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0852113584715103e-05,
+      "loss": 0.0567,
+      "step": 168
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0752067027297486e-05,
+      "loss": 0.0722,
+      "step": 169
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.065194470569193e-05,
+      "loss": 0.0419,
+      "step": 170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0551756706348331e-05,
+      "loss": 0.04,
+      "step": 171
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0451513122333042e-05,
+      "loss": 0.0227,
+      "step": 172
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.035122405231209e-05,
+      "loss": 0.0136,
+      "step": 173
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0250899599533833e-05,
+      "loss": 0.0613,
+      "step": 174
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0150549870811108e-05,
+      "loss": 0.02,
+      "step": 175
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0050184975503104e-05,
+      "loss": 0.0232,
+      "step": 176
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.949815024496901e-06,
+      "loss": 0.0229,
+      "step": 177
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.849450129188895e-06,
+      "loss": 0.0483,
+      "step": 178
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.74910040046617e-06,
+      "loss": 0.0153,
+      "step": 179
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.648775947687914e-06,
+      "loss": 0.0121,
+      "step": 180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.548486877666963e-06,
+      "loss": 0.0603,
+      "step": 181
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.448243293651676e-06,
+      "loss": 0.0291,
+      "step": 182
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.348055294308074e-06,
+      "loss": 0.0458,
+      "step": 183
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.247932972702514e-06,
+      "loss": 0.0438,
+      "step": 184
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.147886415284903e-06,
+      "loss": 0.0337,
+      "step": 185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.047925700872552e-06,
+      "loss": 0.0527,
+      "step": 186
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.948060899634846e-06,
+      "loss": 0.0457,
+      "step": 187
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.848302072078762e-06,
+      "loss": 0.0354,
+      "step": 188
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.748659268035339e-06,
+      "loss": 0.0153,
+      "step": 189
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.649142525647271e-06,
+      "loss": 0.1229,
+      "step": 190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.549761870357633e-06,
+      "loss": 0.0149,
+      "step": 191
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.450527313899923e-06,
+      "loss": 0.042,
+      "step": 192
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.351448853289448e-06,
+      "loss": 0.0146,
+      "step": 193
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.25253646981622e-06,
+      "loss": 0.0145,
+      "step": 194
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.153800128039441e-06,
+      "loss": 0.0691,
+      "step": 195
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.05524977478364e-06,
+      "loss": 0.0165,
+      "step": 196
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.956895338136618e-06,
+      "loss": 0.0776,
+      "step": 197
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.858746726449309e-06,
+      "loss": 0.046,
+      "step": 198
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.760813827337555e-06,
+      "loss": 0.0511,
+      "step": 199
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.663106506686057e-06,
+      "loss": 0.056,
+      "step": 200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.565634607654453e-06,
+      "loss": 0.0084,
+      "step": 201
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.468407949685695e-06,
+      "loss": 0.0492,
+      "step": 202
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.371436327516854e-06,
+      "loss": 0.0472,
+      "step": 203
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.274729510192367e-06,
+      "loss": 0.0826,
+      "step": 204
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.1782972400798825e-06,
+      "loss": 0.062,
+      "step": 205
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.082149231888833e-06,
+      "loss": 0.0373,
+      "step": 206
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.986295171691727e-06,
+      "loss": 0.0227,
+      "step": 207
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.890744715948388e-06,
+      "loss": 0.029,
+      "step": 208
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.795507490533142e-06,
+      "loss": 0.0606,
+      "step": 209
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.700593089765086e-06,
+      "loss": 0.0784,
+      "step": 210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.606011075441556e-06,
+      "loss": 0.0254,
+      "step": 211
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.511770975874862e-06,
+      "loss": 0.0669,
+      "step": 212
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.417882284932373e-06,
+      "loss": 0.0754,
+      "step": 213
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.324354461080121e-06,
+      "loss": 0.0171,
+      "step": 214
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.231196926429913e-06,
+      "loss": 0.054,
+      "step": 215
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.138419065790169e-06,
+      "loss": 0.0068,
+      "step": 216
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.046030225720456e-06,
+      "loss": 0.0406,
+      "step": 217
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.95403971358991e-06,
+      "loss": 0.0413,
+      "step": 218
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.86245679663962e-06,
+      "loss": 0.0359,
+      "step": 219
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.7712907010490036e-06,
+      "loss": 0.0398,
+      "step": 220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.680550611006372e-06,
+      "loss": 0.0679,
+      "step": 221
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.590245667783701e-06,
+      "loss": 0.0356,
+      "step": 222
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.5003849688157075e-06,
+      "loss": 0.0169,
+      "step": 223
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.4109775667833866e-06,
+      "loss": 0.0432,
+      "step": 224
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.322032468702037e-06,
+      "loss": 0.012,
+      "step": 225
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.233558635013842e-06,
+      "loss": 0.0554,
+      "step": 226
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 5.145564978685234e-06,
+      "loss": 0.065,
+      "step": 227
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 5.058060364308965e-06,
+      "loss": 0.0139,
+      "step": 228
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.971053607211069e-06,
+      "loss": 0.0117,
+      "step": 229
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.884553472562809e-06,
+      "loss": 0.0763,
+      "step": 230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.7985686744976714e-06,
+      "loss": 0.0341,
+      "step": 231
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.713107875233459e-06,
+      "loss": 0.0602,
+      "step": 232
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.628179684199685e-06,
+      "loss": 0.0102,
+      "step": 233
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.543792657170228e-06,
+      "loss": 0.0552,
+      "step": 234
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.459955295401415e-06,
+      "loss": 0.0356,
+      "step": 235
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.376676044775601e-06,
+      "loss": 0.0439,
+      "step": 236
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.293963294950313e-06,
+      "loss": 0.0109,
+      "step": 237
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.211825378513066e-06,
+      "loss": 0.0224,
+      "step": 238
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.130270570141931e-06,
+      "loss": 0.0378,
+      "step": 239
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.0493070857719305e-06,
+      "loss": 0.0714,
+      "step": 240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.968943081767358e-06,
+      "loss": 0.0165,
+      "step": 241
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.889186654100089e-06,
+      "loss": 0.0637,
+      "step": 242
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.81004583753399e-06,
+      "loss": 0.0066,
+      "step": 243
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.7315286048154862e-06,
+      "loss": 0.0178,
+      "step": 244
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.6536428658703594e-06,
+      "loss": 0.0485,
+      "step": 245
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.576396467006925e-06,
+      "loss": 0.0283,
+      "step": 246
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.4997971901255588e-06,
+      "loss": 0.0694,
+      "step": 247
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.4238527519347353e-06,
+      "loss": 0.0552,
+      "step": 248
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.3485708031736698e-06,
+      "loss": 0.0436,
+      "step": 249
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.2739589278415252e-06,
+      "loss": 0.0881,
+      "step": 250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.2000246424334315e-06,
+      "loss": 0.039,
+      "step": 251
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.1267753951832523e-06,
+      "loss": 0.0411,
+      "step": 252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.0542185653132216e-06,
+      "loss": 0.0676,
+      "step": 253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.982361462290575e-06,
+      "loss": 0.0206,
+      "step": 254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.9112113250911844e-06,
+      "loss": 0.0827,
+      "step": 255
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.8407753214702694e-06,
+      "loss": 0.0102,
+      "step": 256
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7710605472403373e-06,
+      "loss": 0.0229,
+      "step": 257
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.702074025556327e-06,
+      "loss": 0.0196,
+      "step": 258
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6338227062080924e-06,
+      "loss": 0.0625,
+      "step": 259
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.566313464920265e-06,
+      "loss": 0.0595,
+      "step": 260
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4995531026595952e-06,
+      "loss": 0.0147,
+      "step": 261
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4335483449498053e-06,
+      "loss": 0.0544,
+      "step": 262
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3683058411940563e-06,
+      "loss": 0.0453,
+      "step": 263
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3038321640050763e-06,
+      "loss": 0.0609,
+      "step": 264
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.2401338085430326e-06,
+      "loss": 0.0504,
+      "step": 265
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.177217191861183e-06,
+      "loss": 0.0248,
+      "step": 266
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.115088652259446e-06,
+      "loss": 0.0616,
+      "step": 267
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.053754448645846e-06,
+      "loss": 0.0408,
+      "step": 268
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.9932207599059782e-06,
+      "loss": 0.0444,
+      "step": 269
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.933493684280574e-06,
+      "loss": 0.0632,
+      "step": 270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8745792387511241e-06,
+      "loss": 0.019,
+      "step": 271
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8164833584337216e-06,
+      "loss": 0.0266,
+      "step": 272
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.75921189598118e-06,
+      "loss": 0.0499,
+      "step": 273
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7027706209933903e-06,
+      "loss": 0.0379,
+      "step": 274
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6471652194361131e-06,
+      "loss": 0.0092,
+      "step": 275
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5924012930681643e-06,
+      "loss": 0.0326,
+      "step": 276
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5384843588770626e-06,
+      "loss": 0.0179,
+      "step": 277
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4854198485232696e-06,
+      "loss": 0.0641,
+      "step": 278
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433213107792991e-06,
+      "loss": 0.0285,
+      "step": 279
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.3818693960596186e-06,
+      "loss": 0.0365,
+      "step": 280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.3313938857539133e-06,
+      "loss": 0.0873,
+      "step": 281
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2817916618429194e-06,
+      "loss": 0.0148,
+      "step": 282
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.2330677213177034e-06,
+      "loss": 0.0127,
+      "step": 283
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1852269726899423e-06,
+      "loss": 0.0028,
+      "step": 284
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.138274235497443e-06,
+      "loss": 0.0241,
+      "step": 285
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0922142398186097e-06,
+      "loss": 0.0466,
+      "step": 286
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0470516257959351e-06,
+      "loss": 0.0273,
+      "step": 287
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.00279094316854e-06,
+      "loss": 0.0529,
+      "step": 288
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.594366508138352e-07,
+      "loss": 0.0648,
+      "step": 289
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.169931162983137e-07,
+      "loss": 0.0118,
+      "step": 290
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.754646154375801e-07,
+      "loss": 0.0321,
+      "step": 291
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.348553318655795e-07,
+      "loss": 0.0167,
+      "step": 292
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.951693566131325e-07,
+      "loss": 0.0204,
+      "step": 293
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.564106876958188e-07,
+      "loss": 0.0502,
+      "step": 294
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.185832297111939e-07,
+      "loss": 0.0075,
+      "step": 295
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.816907934454353e-07,
+      "loss": 0.053,
+      "step": 296
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.457370954894582e-07,
+      "loss": 0.0255,
+      "step": 297
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.107257578644721e-07,
+      "loss": 0.0422,
+      "step": 298
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.766603076571164e-07,
+      "loss": 0.0745,
+      "step": 299
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.43544176664137e-07,
+      "loss": 0.0217,
+      "step": 300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.113807010466432e-07,
+      "loss": 0.0417,
+      "step": 301
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.801731209940375e-07,
+      "loss": 0.0324,
+      "step": 302
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.499245803975927e-07,
+      "loss": 0.0264,
+      "step": 303
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.206381265337189e-07,
+      "loss": 0.0285,
+      "step": 304
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.9231670975699354e-07,
+      "loss": 0.0603,
+      "step": 305
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.649631832029288e-07,
+      "loss": 0.0515,
+      "step": 306
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.385803025005463e-07,
+      "loss": 0.0133,
+      "step": 307
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.1317072549477246e-07,
+      "loss": 0.0106,
+      "step": 308
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.887370119786792e-07,
+      "loss": 0.0492,
+      "step": 309
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6528162343561593e-07,
+      "loss": 0.0132,
+      "step": 310
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4280692279122554e-07,
+      "loss": 0.0207,
+      "step": 311
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2131517417540937e-07,
+      "loss": 0.0301,
+      "step": 312
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.00808542694233e-07,
+      "loss": 0.0294,
+      "step": 313
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8128909421180506e-07,
+      "loss": 0.0349,
+      "step": 314
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6275879514217052e-07,
+      "loss": 0.0405,
+      "step": 315
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4521951225120345e-07,
+      "loss": 0.0464,
+      "step": 316
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2867301246854757e-07,
+      "loss": 0.0258,
+      "step": 317
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1312096270961525e-07,
+      "loss": 0.0217,
+      "step": 318
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.856492970766296e-08,
+      "loss": 0.0521,
+      "step": 319
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.50063798559475e-08,
+      "loss": 0.074,
+      "step": 320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.244667906001202e-08,
+      "loss": 0.0091,
+      "step": 321
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.088709260007153e-08,
+      "loss": 0.0584,
+      "step": 322
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.032878500355498e-08,
+      "loss": 0.0114,
+      "step": 323
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.07728199277857e-08,
+      "loss": 0.0528,
+      "step": 324
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2220160052828245e-08,
+      "loss": 0.0534,
+      "step": 325
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.467166698450485e-08,
+      "loss": 0.0912,
+      "step": 326
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.812810116760044e-08,
+      "loss": 0.0426,
+      "step": 327
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2590121809247235e-08,
+      "loss": 0.0501,
+      "step": 328
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.05828681252452e-09,
+      "loss": 0.0135,
+      "step": 329
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.5330527202480656e-09,
+      "loss": 0.0375,
+      "step": 330
+    },
+    {
+      "epoch": 1.0,
+      "step": 330,
+      "total_flos": 1006103224320.0,
+      "train_loss": 0.05386477595011732,
+      "train_runtime": 2481.834,
+      "train_samples_per_second": 2.652,
+      "train_steps_per_second": 0.133
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 330,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "total_flos": 1006103224320.0,
+  "train_batch_size": 10,
+  "trial_name": null,
+  "trial_params": null
+}

CheckGuard Models/wholeimage/bank/finetune_lora_llava_mistral.sh

@@ -0,0 +1,43 @@
+#!/bin/bash
+# Use first parameter as GPU IDs, default to "0,1,2,3" if not provided
+GPU_IDS=${1:-0,1,2,3}
+
+
+CUDA_VISIBLE_DEVICES=0,1,2,3 deepspeed --include localhost:"$GPU_IDS" --master_port 29601\
+    llava/train/train_mem.py \
+    --lora_enable True --lora_r 128 --lora_alpha 256 --mm_projector_lr 2e-5 \
+    --deepspeed ./scripts/zero3.json \
+    --model_name_or_path liuhaotian/llava-v1.6-mistral-7b \
+    --version mistral_instruct \
+    --data_path /home/larry5/project/LLaVA-1.6-ft/data/peft/bank/bank_dataset.json \
+    --image_folder /home/larry5/project/LLaVA-1.6-ft/data/data/ \
+    --vision_tower openai/clip-vit-large-patch14-336 \
+    --mm_projector_type mlp2x_gelu \
+    --mm_vision_select_layer -2 \
+    --mm_use_im_start_end False \
+    --mm_use_im_patch_token False \
+    --mm_patch_merge_type spatial_unpad \
+    --image_aspect_ratio anyres \
+    --group_by_modality_length False \
+    --bf16 False \
+    --fp16 True \
+    --output_dir /home/larry5/project/LLaVA-1.6-ft/scripts_peft/mistral/lora/llava-lora-mistral-r128a256/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model \
+    --num_train_epochs 1 \
+    --per_device_train_batch_size 10 \
+    --per_device_eval_batch_size 1 \
+    --gradient_accumulation_steps 1 \
+    --evaluation_strategy "no" \
+    --save_strategy "steps" \
+    --save_steps 500 \
+    --save_total_limit 5 \
+    --learning_rate 2e-5 \
+    --weight_decay 0. \
+    --warmup_ratio 0.05 \
+    --lr_scheduler_type "cosine" \
+    --logging_steps 1 \
+    --tf32 True \
+    --model_max_length 4096 \
+    --gradient_checkpointing True \
+    --dataloader_num_workers 4 \
+    --lazy_preprocess True \
+    --report_to wandb \

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/README.md

@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: liuhaotian/llava-v1.6-mistral-7b
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.10.0

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "liuhaotian/llava-v1.6-mistral-7b",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "v_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "k_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b234ad29295da4f261427006e770781d152d27e2bd090a65ac32cfb8472dd11b
+size 708924928

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/README.md

@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: liuhaotian/llava-v1.6-mistral-7b
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.10.0

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "liuhaotian/llava-v1.6-mistral-7b",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "v_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "k_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c26e454370836d2d8ee1827620ba4d532f3b135ebd1e7fcbb0263a086f241253
+size 1417762896

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/global_step500/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3183212032ce3f53bf011c0ea2d72e73d90e4ae83d758f3cb2661945c405d2e
+size 632242

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/global_step500/zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59f2e280a5a1fb1c30380e867fb4625232d56bf1130fb2ac1bda1c76272752dd
+size 4504787266

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/latest

@@ -0,0 +1 @@
+global_step500

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42d7cdbb5673ea29475539a9e027f8b9828b8bdf3f8f5a3383b13244fc3604a3
+size 14244

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74f73b67322f406ba2e53b1ed170e4b3c50a5de49d1b4aa38bda0b32a3724ada
+size 1064

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/special_tokens_map.json

@@ -0,0 +1,24 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<unk>",
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055
+size 493443

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/tokenizer_config.json

@@ -0,0 +1,44 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [],
+  "bos_token": "<s>",
+  "chat_template": "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token}}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": true,
+  "model_max_length": 4096,
+  "pad_token": "<unk>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/trainer_state.json

@@ -0,0 +1,3021 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9009009009009009,
+  "eval_steps": 500,
+  "global_step": 500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.142857142857143e-07,
+      "loss": 0.4237,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4285714285714286e-06,
+      "loss": 0.3368,
+      "step": 2
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1428571428571427e-06,
+      "loss": 0.214,
+      "step": 3
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8571428571428573e-06,
+      "loss": 0.396,
+      "step": 4
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5714285714285718e-06,
+      "loss": 0.305,
+      "step": 5
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2857142857142855e-06,
+      "loss": 0.4049,
+      "step": 6
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5e-06,
+      "loss": 0.108,
+      "step": 7
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.7142857142857145e-06,
+      "loss": 0.2286,
+      "step": 8
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.4285714285714295e-06,
+      "loss": 0.1443,
+      "step": 9
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.1428571428571436e-06,
+      "loss": 0.2252,
+      "step": 10
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.857142857142858e-06,
+      "loss": 0.0747,
+      "step": 11
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 0.1084,
+      "step": 12
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.285714285714288e-06,
+      "loss": 0.2115,
+      "step": 13
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-05,
+      "loss": 0.4742,
+      "step": 14
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0714285714285714e-05,
+      "loss": 0.083,
+      "step": 15
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 0.3392,
+      "step": 16
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.2142857142857142e-05,
+      "loss": 0.065,
+      "step": 17
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.2857142857142859e-05,
+      "loss": 0.1711,
+      "step": 18
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3571428571428574e-05,
+      "loss": 0.0539,
+      "step": 19
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4285714285714287e-05,
+      "loss": 0.0701,
+      "step": 20
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.0836,
+      "step": 21
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.5714285714285715e-05,
+      "loss": 0.1891,
+      "step": 22
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.642857142857143e-05,
+      "loss": 0.0422,
+      "step": 23
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.7142857142857142e-05,
+      "loss": 0.2094,
+      "step": 24
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.785714285714286e-05,
+      "loss": 0.139,
+      "step": 25
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8571428571428575e-05,
+      "loss": 0.2214,
+      "step": 26
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.928571428571429e-05,
+      "loss": 0.1084,
+      "step": 27
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2e-05,
+      "loss": 0.0898,
+      "step": 28
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9999822316445652e-05,
+      "loss": 0.0359,
+      "step": 29
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9999289272096886e-05,
+      "loss": 0.2648,
+      "step": 30
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9998400885896355e-05,
+      "loss": 0.4007,
+      "step": 31
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9997157189414373e-05,
+      "loss": 0.235,
+      "step": 32
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999555822684783e-05,
+      "loss": 0.0273,
+      "step": 33
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999360405501859e-05,
+      "loss": 0.0267,
+      "step": 34
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99912947433715e-05,
+      "loss": 0.2619,
+      "step": 35
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9988630373971896e-05,
+      "loss": 0.4101,
+      "step": 36
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9985611041502704e-05,
+      "loss": 0.1302,
+      "step": 37
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9982236853261067e-05,
+      "loss": 0.118,
+      "step": 38
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9978507929154534e-05,
+      "loss": 0.0933,
+      "step": 39
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997442440169681e-05,
+      "loss": 0.0104,
+      "step": 40
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9969986416003026e-05,
+      "loss": 0.1061,
+      "step": 41
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965194129784597e-05,
+      "loss": 0.1575,
+      "step": 42
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996004771334361e-05,
+      "loss": 0.1969,
+      "step": 43
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996004771334361e-05,
+      "loss": 0.0492,
+      "step": 44
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9954547349566783e-05,
+      "loss": 0.3012,
+      "step": 45
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.994869323391895e-05,
+      "loss": 0.2185,
+      "step": 46
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.994248557443613e-05,
+      "loss": 0.1729,
+      "step": 47
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.993592459171812e-05,
+      "loss": 0.0354,
+      "step": 48
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9929010518920667e-05,
+      "loss": 0.3939,
+      "step": 49
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.992174360174717e-05,
+      "loss": 0.0505,
+      "step": 50
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9914124098439976e-05,
+      "loss": 0.0777,
+      "step": 51
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9914124098439976e-05,
+      "loss": 0.6651,
+      "step": 52
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9906152279771162e-05,
+      "loss": 0.15,
+      "step": 53
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9897828429032946e-05,
+      "loss": 0.1416,
+      "step": 54
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9889152842027607e-05,
+      "loss": 0.1195,
+      "step": 55
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9880125827056967e-05,
+      "loss": 0.0787,
+      "step": 56
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.987074770491145e-05,
+      "loss": 0.0681,
+      "step": 57
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.986101880885867e-05,
+      "loss": 0.1337,
+      "step": 58
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9850939484631598e-05,
+      "loss": 0.0961,
+      "step": 59
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.984051009041626e-05,
+      "loss": 0.116,
+      "step": 60
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.982973099683902e-05,
+      "loss": 0.3853,
+      "step": 61
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9818602586953414e-05,
+      "loss": 0.0875,
+      "step": 62
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9807125256226532e-05,
+      "loss": 0.3216,
+      "step": 63
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9795299412524948e-05,
+      "loss": 0.0752,
+      "step": 64
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9783125476100254e-05,
+      "loss": 0.1461,
+      "step": 65
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9770603879574108e-05,
+      "loss": 0.075,
+      "step": 66
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.975773506792287e-05,
+      "loss": 0.0685,
+      "step": 67
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.974451949846177e-05,
+      "loss": 0.0555,
+      "step": 68
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.973095764082869e-05,
+      "loss": 0.0171,
+      "step": 69
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9717049976967437e-05,
+      "loss": 0.0247,
+      "step": 70
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9702797001110642e-05,
+      "loss": 0.0839,
+      "step": 71
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9688199219762183e-05,
+      "loss": 0.4163,
+      "step": 72
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.96732571516792e-05,
+      "loss": 0.1461,
+      "step": 73
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9657971327853644e-05,
+      "loss": 0.1457,
+      "step": 74
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.964234229149342e-05,
+      "loss": 0.0482,
+      "step": 75
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.962637059800307e-05,
+      "loss": 0.0802,
+      "step": 76
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9610056814964053e-05,
+      "loss": 0.0697,
+      "step": 77
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.959340152211455e-05,
+      "loss": 0.0614,
+      "step": 78
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.95764053113289e-05,
+      "loss": 0.1004,
+      "step": 79
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9559068786596526e-05,
+      "loss": 0.0286,
+      "step": 80
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.954139256400049e-05,
+      "loss": 0.1162,
+      "step": 81
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.952337727169561e-05,
+      "loss": 0.0731,
+      "step": 82
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.950502354988612e-05,
+      "loss": 0.0286,
+      "step": 83
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.948633205080292e-05,
+      "loss": 0.2425,
+      "step": 84
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9467303438680414e-05,
+      "loss": 0.0505,
+      "step": 85
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.944793838973289e-05,
+      "loss": 0.0922,
+      "step": 86
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9428237592130487e-05,
+      "loss": 0.2949,
+      "step": 87
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.940820174597476e-05,
+      "loss": 0.2807,
+      "step": 88
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9387831563273775e-05,
+      "loss": 0.2377,
+      "step": 89
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9367127767916828e-05,
+      "loss": 0.2558,
+      "step": 90
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9346091095648712e-05,
+      "loss": 0.0871,
+      "step": 91
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.932472229404356e-05,
+      "loss": 0.2204,
+      "step": 92
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9303022122478303e-05,
+      "loss": 0.1174,
+      "step": 93
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9280991352105656e-05,
+      "loss": 0.2181,
+      "step": 94
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.925863076582674e-05,
+      "loss": 0.1251,
+      "step": 95
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9235941158263253e-05,
+      "loss": 0.2251,
+      "step": 96
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9212923335729206e-05,
+      "loss": 0.1236,
+      "step": 97
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.918957811620231e-05,
+      "loss": 0.0901,
+      "step": 98
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9165906329294875e-05,
+      "loss": 0.1002,
+      "step": 99
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9141908816224356e-05,
+      "loss": 0.4397,
+      "step": 100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9117586429783433e-05,
+      "loss": 0.1141,
+      "step": 101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.909294003430972e-05,
+      "loss": 0.1842,
+      "step": 102
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.906797050565505e-05,
+      "loss": 0.0985,
+      "step": 103
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9042678731154337e-05,
+      "loss": 0.1533,
+      "step": 104
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.901706560959407e-05,
+      "loss": 0.145,
+      "step": 105
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8991132051180332e-05,
+      "loss": 0.1693,
+      "step": 106
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8964878977506496e-05,
+      "loss": 0.2012,
+      "step": 107
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8938307321520453e-05,
+      "loss": 0.1286,
+      "step": 108
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8911418027491453e-05,
+      "loss": 0.1396,
+      "step": 109
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8884212050976568e-05,
+      "loss": 0.0291,
+      "step": 110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.885669035878672e-05,
+      "loss": 0.0317,
+      "step": 111
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.882885392895232e-05,
+      "loss": 0.1143,
+      "step": 112
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8800703750688536e-05,
+      "loss": 0.126,
+      "step": 113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.877224082436011e-05,
+      "loss": 0.2017,
+      "step": 114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8743466161445823e-05,
+      "loss": 0.0735,
+      "step": 115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8714380784502553e-05,
+      "loss": 0.0527,
+      "step": 116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8684985727128936e-05,
+      "loss": 0.1112,
+      "step": 117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8655282033928618e-05,
+      "loss": 0.3129,
+      "step": 118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8625270760473164e-05,
+      "loss": 0.2827,
+      "step": 119
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8594952973264512e-05,
+      "loss": 0.5608,
+      "step": 120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856432974969711e-05,
+      "loss": 0.1465,
+      "step": 121
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8533402178019596e-05,
+      "loss": 0.1322,
+      "step": 122
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8502171357296144e-05,
+      "loss": 0.0912,
+      "step": 123
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8470638397367397e-05,
+      "loss": 0.0419,
+      "step": 124
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8438804418811038e-05,
+      "loss": 0.0369,
+      "step": 125
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8406670552901958e-05,
+      "loss": 0.0529,
+      "step": 126
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.837423794157206e-05,
+      "loss": 0.1472,
+      "step": 127
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.834150773736967e-05,
+      "loss": 0.0425,
+      "step": 128
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8308481103418597e-05,
+      "loss": 0.1634,
+      "step": 129
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8275159213376783e-05,
+      "loss": 0.0485,
+      "step": 130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.82415432513946e-05,
+      "loss": 0.0313,
+      "step": 131
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8207634412072765e-05,
+      "loss": 0.1792,
+      "step": 132
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.81734339004199e-05,
+      "loss": 0.1184,
+      "step": 133
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8138942931809702e-05,
+      "loss": 0.2756,
+      "step": 134
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8104162731937746e-05,
+      "loss": 0.0635,
+      "step": 135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8069094536777938e-05,
+      "loss": 0.0158,
+      "step": 136
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8033739592538598e-05,
+      "loss": 0.2732,
+      "step": 137
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7998099155618147e-05,
+      "loss": 0.1428,
+      "step": 138
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7962174492560492e-05,
+      "loss": 0.0777,
+      "step": 139
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7925966880009998e-05,
+      "loss": 0.1644,
+      "step": 140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7889477604666124e-05,
+      "loss": 0.0999,
+      "step": 141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.785270796323769e-05,
+      "loss": 0.0446,
+      "step": 142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7815659262396825e-05,
+      "loss": 0.0647,
+      "step": 143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7778332818732492e-05,
+      "loss": 0.0521,
+      "step": 144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7740729958703725e-05,
+      "loss": 0.2041,
+      "step": 145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7702852018592493e-05,
+      "loss": 0.0149,
+      "step": 146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7664700344456198e-05,
+      "loss": 0.0502,
+      "step": 147
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.762627629207986e-05,
+      "loss": 0.2027,
+      "step": 148
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.758758122692791e-05,
+      "loss": 0.0187,
+      "step": 149
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7548616524095697e-05,
+      "loss": 0.1248,
+      "step": 150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7509383568260597e-05,
+      "loss": 0.0859,
+      "step": 151
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7469883753632817e-05,
+      "loss": 0.0822,
+      "step": 152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.743011848390585e-05,
+      "loss": 0.2445,
+      "step": 153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7390089172206594e-05,
+      "loss": 0.2662,
+      "step": 154
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7349797241045115e-05,
+      "loss": 0.0984,
+      "step": 155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.730924412226413e-05,
+      "loss": 0.0317,
+      "step": 156
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.726843125698809e-05,
+      "loss": 0.1129,
+      "step": 157
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7227360095571992e-05,
+      "loss": 0.1882,
+      "step": 158
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7186032097549822e-05,
+      "loss": 0.1099,
+      "step": 159
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7144448731582698e-05,
+      "loss": 0.3506,
+      "step": 160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7102611475406676e-05,
+      "loss": 0.0936,
+      "step": 161
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7060521815780225e-05,
+      "loss": 0.104,
+      "step": 162
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7018181248431416e-05,
+      "loss": 0.168,
+      "step": 163
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6975591278004747e-05,
+      "loss": 0.2726,
+      "step": 164
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6932753418007683e-05,
+      "loss": 0.0564,
+      "step": 165
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.688966919075687e-05,
+      "loss": 0.2981,
+      "step": 166
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.684634012732403e-05,
+      "loss": 0.0602,
+      "step": 167
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.680276776748157e-05,
+      "loss": 0.0364,
+      "step": 168
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6758953659647838e-05,
+      "loss": 0.096,
+      "step": 169
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6714899360832118e-05,
+      "loss": 0.2139,
+      "step": 170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.667060643657929e-05,
+      "loss": 0.1666,
+      "step": 171
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.66260764609142e-05,
+      "loss": 0.0486,
+      "step": 172
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.658131101628571e-05,
+      "loss": 0.055,
+      "step": 173
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.653631169351049e-05,
+      "loss": 0.0953,
+      "step": 174
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6491080091716457e-05,
+      "loss": 0.1824,
+      "step": 175
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6445617818285974e-05,
+      "loss": 0.0226,
+      "step": 176
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6399926488798702e-05,
+      "loss": 0.0388,
+      "step": 177
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6354007726974205e-05,
+      "loss": 0.1149,
+      "step": 178
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.630786316461425e-05,
+      "loss": 0.1428,
+      "step": 179
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6261494441544805e-05,
+      "loss": 0.0445,
+      "step": 180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6214903205557774e-05,
+      "loss": 0.0612,
+      "step": 181
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6168091112352443e-05,
+      "loss": 0.0826,
+      "step": 182
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.612105982547663e-05,
+      "loss": 0.0376,
+      "step": 183
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.607381101626758e-05,
+      "loss": 0.1441,
+      "step": 184
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6026346363792565e-05,
+      "loss": 0.1089,
+      "step": 185
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5978667554789216e-05,
+      "loss": 0.0845,
+      "step": 186
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5930776283605585e-05,
+      "loss": 0.0835,
+      "step": 187
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5882674252139928e-05,
+      "loss": 0.0762,
+      "step": 188
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5834363169780227e-05,
+      "loss": 0.067,
+      "step": 189
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.578584475334345e-05,
+      "loss": 0.0327,
+      "step": 190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5737120727014535e-05,
+      "loss": 0.0254,
+      "step": 191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5688192822285116e-05,
+      "loss": 0.028,
+      "step": 192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5639062777892e-05,
+      "loss": 0.1708,
+      "step": 193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5589732339755362e-05,
+      "loss": 0.0542,
+      "step": 194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5540203260916728e-05,
+      "loss": 0.2358,
+      "step": 195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5490477301476648e-05,
+      "loss": 0.1471,
+      "step": 196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5440556228532168e-05,
+      "loss": 0.0414,
+      "step": 197
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5390441816114022e-05,
+      "loss": 0.0754,
+      "step": 198
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.534013584512359e-05,
+      "loss": 0.105,
+      "step": 199
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5289640103269626e-05,
+      "loss": 0.2052,
+      "step": 200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5238956385004703e-05,
+      "loss": 0.2482,
+      "step": 201
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5188086491461467e-05,
+      "loss": 0.0967,
+      "step": 202
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5137032230388613e-05,
+      "loss": 0.1314,
+      "step": 203
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5085795416086655e-05,
+      "loss": 0.2313,
+      "step": 204
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5034377869343453e-05,
+      "loss": 0.1304,
+      "step": 205
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4982781417369496e-05,
+      "loss": 0.2304,
+      "step": 206
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4931007893732981e-05,
+      "loss": 0.0508,
+      "step": 207
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4879059138294647e-05,
+      "loss": 0.1389,
+      "step": 208
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4826936997142399e-05,
+      "loss": 0.2129,
+      "step": 209
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4774643322525691e-05,
+      "loss": 0.0201,
+      "step": 210
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4722179972789725e-05,
+      "loss": 0.1064,
+      "step": 211
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.466954881230939e-05,
+      "loss": 0.0459,
+      "step": 212
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4616751711423016e-05,
+      "loss": 0.2229,
+      "step": 213
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4563790546365914e-05,
+      "loss": 0.1464,
+      "step": 214
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4510667199203697e-05,
+      "loss": 0.0558,
+      "step": 215
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4457383557765385e-05,
+      "loss": 0.0214,
+      "step": 216
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4403941515576344e-05,
+      "loss": 0.1551,
+      "step": 217
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4350342971790979e-05,
+      "loss": 0.2093,
+      "step": 218
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4296589831125234e-05,
+      "loss": 0.0453,
+      "step": 219
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4242684003788934e-05,
+      "loss": 0.0317,
+      "step": 220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.418862740541788e-05,
+      "loss": 0.1334,
+      "step": 221
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4134421957005775e-05,
+      "loss": 0.0185,
+      "step": 222
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4080069584835971e-05,
+      "loss": 0.087,
+      "step": 223
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4025572220412998e-05,
+      "loss": 0.1747,
+      "step": 224
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3970931800393943e-05,
+      "loss": 0.1168,
+      "step": 225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.391615026651961e-05,
+      "loss": 0.5095,
+      "step": 226
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3861229565545532e-05,
+      "loss": 0.1157,
+      "step": 227
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3806171649172782e-05,
+      "loss": 0.1201,
+      "step": 228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3750978473978611e-05,
+      "loss": 0.2232,
+      "step": 229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3695652001346928e-05,
+      "loss": 0.1718,
+      "step": 230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.36401941973986e-05,
+      "loss": 0.0509,
+      "step": 231
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3584607032921566e-05,
+      "loss": 0.0333,
+      "step": 232
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3528892483300821e-05,
+      "loss": 0.1811,
+      "step": 233
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3473052528448203e-05,
+      "loss": 0.1771,
+      "step": 234
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3417089152732049e-05,
+      "loss": 0.1098,
+      "step": 235
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3361004344906652e-05,
+      "loss": 0.0566,
+      "step": 236
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.330480009804162e-05,
+      "loss": 0.2864,
+      "step": 237
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3248478409451017e-05,
+      "loss": 0.0166,
+      "step": 238
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3192041280622409e-05,
+      "loss": 0.2239,
+      "step": 239
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3135490717145726e-05,
+      "loss": 0.2247,
+      "step": 240
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3078828728641994e-05,
+      "loss": 0.1758,
+      "step": 241
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.3022057328691915e-05,
+      "loss": 0.0618,
+      "step": 242
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2965178534764311e-05,
+      "loss": 0.1204,
+      "step": 243
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2908194368144437e-05,
+      "loss": 0.0233,
+      "step": 244
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.285110685386215e-05,
+      "loss": 0.0387,
+      "step": 245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2793918020619937e-05,
+      "loss": 0.0791,
+      "step": 246
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2736629900720832e-05,
+      "loss": 0.0106,
+      "step": 247
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2679244529996182e-05,
+      "loss": 0.042,
+      "step": 248
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.262176394773332e-05,
+      "loss": 0.0725,
+      "step": 249
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.256419019660308e-05,
+      "loss": 0.0834,
+      "step": 250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2506525322587207e-05,
+      "loss": 0.0432,
+      "step": 251
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2448771374905655e-05,
+      "loss": 0.177,
+      "step": 252
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2390930405943766e-05,
+      "loss": 0.0887,
+      "step": 253
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.233300447117933e-05,
+      "loss": 0.0152,
+      "step": 254
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2274995629109545e-05,
+      "loss": 0.0317,
+      "step": 255
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2216905941177854e-05,
+      "loss": 0.0268,
+      "step": 256
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.215873747170071e-05,
+      "loss": 0.1685,
+      "step": 257
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2100492287794186e-05,
+      "loss": 0.1403,
+      "step": 258
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.2042172459300546e-05,
+      "loss": 0.0443,
+      "step": 259
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.198378005871467e-05,
+      "loss": 0.3589,
+      "step": 260
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.192531716111042e-05,
+      "loss": 0.0427,
+      "step": 261
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1866785844066884e-05,
+      "loss": 0.1103,
+      "step": 262
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1808188187594549e-05,
+      "loss": 0.2563,
+      "step": 263
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1749526274061394e-05,
+      "loss": 0.1494,
+      "step": 264
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1690802188118878e-05,
+      "loss": 0.1105,
+      "step": 265
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1632018016627859e-05,
+      "loss": 0.082,
+      "step": 266
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1573175848584455e-05,
+      "loss": 0.3555,
+      "step": 267
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1514277775045768e-05,
+      "loss": 0.0603,
+      "step": 268
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1455325889055616e-05,
+      "loss": 0.2883,
+      "step": 269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1396322285570119e-05,
+      "loss": 0.054,
+      "step": 270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1337269061383278e-05,
+      "loss": 0.0668,
+      "step": 271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1278168315052445e-05,
+      "loss": 0.1454,
+      "step": 272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1219022146823762e-05,
+      "loss": 0.0619,
+      "step": 273
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1159832658557498e-05,
+      "loss": 0.0449,
+      "step": 274
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1100601953653393e-05,
+      "loss": 0.0684,
+      "step": 275
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1041332136975874e-05,
+      "loss": 0.0273,
+      "step": 276
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0982025314779287e-05,
+      "loss": 0.2375,
+      "step": 277
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.092268359463302e-05,
+      "loss": 0.0353,
+      "step": 278
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.086330908534663e-05,
+      "loss": 0.1224,
+      "step": 279
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0803903896894877e-05,
+      "loss": 0.1297,
+      "step": 280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0744470140342775e-05,
+      "loss": 0.4464,
+      "step": 281
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0685009927770542e-05,
+      "loss": 0.103,
+      "step": 282
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0625525372198564e-05,
+      "loss": 0.0881,
+      "step": 283
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.056601858751229e-05,
+      "loss": 0.075,
+      "step": 284
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0506491688387128e-05,
+      "loss": 0.0677,
+      "step": 285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0446946790213275e-05,
+      "loss": 0.2301,
+      "step": 286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0387386009020569e-05,
+      "loss": 0.0737,
+      "step": 287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.032781146140326e-05,
+      "loss": 0.1262,
+      "step": 288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0268225264444829e-05,
+      "loss": 0.0252,
+      "step": 289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0208629535642726e-05,
+      "loss": 0.0192,
+      "step": 290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0149026392833137e-05,
+      "loss": 0.257,
+      "step": 291
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0089417954115715e-05,
+      "loss": 0.1876,
+      "step": 292
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.002980633777831e-05,
+      "loss": 0.0341,
+      "step": 293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.970193662221694e-06,
+      "loss": 0.232,
+      "step": 294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.910582045884292e-06,
+      "loss": 0.1429,
+      "step": 295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.850973607166865e-06,
+      "loss": 0.2432,
+      "step": 296
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.791370464357279e-06,
+      "loss": 0.0288,
+      "step": 297
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.731774735555174e-06,
+      "loss": 0.2272,
+      "step": 298
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.672188538596746e-06,
+      "loss": 0.1102,
+      "step": 299
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.612613990979436e-06,
+      "loss": 0.0529,
+      "step": 300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.553053209786725e-06,
+      "loss": 0.1721,
+      "step": 301
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.493508311612874e-06,
+      "loss": 0.0046,
+      "step": 302
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.433981412487711e-06,
+      "loss": 0.043,
+      "step": 303
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.374474627801439e-06,
+      "loss": 0.0589,
+      "step": 304
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.314990072229461e-06,
+      "loss": 0.0114,
+      "step": 305
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.25552985965723e-06,
+      "loss": 0.1645,
+      "step": 306
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.196096103105127e-06,
+      "loss": 0.2002,
+      "step": 307
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.136690914653377e-06,
+      "loss": 0.057,
+      "step": 308
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.07731640536698e-06,
+      "loss": 0.1744,
+      "step": 309
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.017974685220716e-06,
+      "loss": 0.0343,
+      "step": 310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.958667863024127e-06,
+      "loss": 0.0405,
+      "step": 311
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.899398046346608e-06,
+      "loss": 0.2055,
+      "step": 312
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.840167341442505e-06,
+      "loss": 0.0673,
+      "step": 313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.78097785317624e-06,
+      "loss": 0.0291,
+      "step": 314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.721831684947557e-06,
+      "loss": 0.2443,
+      "step": 315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.662730938616724e-06,
+      "loss": 0.058,
+      "step": 316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.603677714429888e-06,
+      "loss": 0.2347,
+      "step": 317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.54467411094439e-06,
+      "loss": 0.0307,
+      "step": 318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.485722224954237e-06,
+      "loss": 0.0094,
+      "step": 319
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.426824151415548e-06,
+      "loss": 0.0724,
+      "step": 320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.367981983372143e-06,
+      "loss": 0.0816,
+      "step": 321
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.309197811881128e-06,
+      "loss": 0.0375,
+      "step": 322
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.250473725938608e-06,
+      "loss": 0.0106,
+      "step": 323
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.191811812405453e-06,
+      "loss": 0.0701,
+      "step": 324
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.133214155933118e-06,
+      "loss": 0.0134,
+      "step": 325
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.074682838889581e-06,
+      "loss": 0.1992,
+      "step": 326
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.01621994128533e-06,
+      "loss": 0.1688,
+      "step": 327
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.95782754069946e-06,
+      "loss": 0.2751,
+      "step": 328
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.899507712205818e-06,
+      "loss": 0.0192,
+      "step": 329
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.841262528299296e-06,
+      "loss": 0.0797,
+      "step": 330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.783094058822147e-06,
+      "loss": 0.0867,
+      "step": 331
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.72500437089046e-06,
+      "loss": 0.0445,
+      "step": 332
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.666995528820673e-06,
+      "loss": 0.1654,
+      "step": 333
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.609069594056234e-06,
+      "loss": 0.0168,
+      "step": 334
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.551228625094349e-06,
+      "loss": 0.0779,
+      "step": 335
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.493474677412795e-06,
+      "loss": 0.0444,
+      "step": 336
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.435809803396923e-06,
+      "loss": 0.1839,
+      "step": 337
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.37823605226668e-06,
+      "loss": 0.3834,
+      "step": 338
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.320755470003822e-06,
+      "loss": 0.0261,
+      "step": 339
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.263370099279173e-06,
+      "loss": 0.0084,
+      "step": 340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.2060819793800665e-06,
+      "loss": 0.0469,
+      "step": 341
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.148893146137852e-06,
+      "loss": 0.3605,
+      "step": 342
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.091805631855566e-06,
+      "loss": 0.0621,
+      "step": 343
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.034821465235693e-06,
+      "loss": 0.099,
+      "step": 344
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.977942671308087e-06,
+      "loss": 0.0641,
+      "step": 345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.921171271358007e-06,
+      "loss": 0.0859,
+      "step": 346
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.864509282854272e-06,
+      "loss": 0.0564,
+      "step": 347
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.8079587193775935e-06,
+      "loss": 0.0405,
+      "step": 348
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.751521590548986e-06,
+      "loss": 0.101,
+      "step": 349
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.695199901958386e-06,
+      "loss": 0.1178,
+      "step": 350
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.638995655093351e-06,
+      "loss": 0.2406,
+      "step": 351
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.582910847267957e-06,
+      "loss": 0.1846,
+      "step": 352
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.526947471551799e-06,
+      "loss": 0.1374,
+      "step": 353
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.471107516699183e-06,
+      "loss": 0.0863,
+      "step": 354
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.415392967078438e-06,
+      "loss": 0.0755,
+      "step": 355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.3598058026013995e-06,
+      "loss": 0.0732,
+      "step": 356
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.304347998653074e-06,
+      "loss": 0.0555,
+      "step": 357
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.24902152602139e-06,
+      "loss": 0.0475,
+      "step": 358
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.193828350827222e-06,
+      "loss": 0.036,
+      "step": 359
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.1387704344544684e-06,
+      "loss": 0.2679,
+      "step": 360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.083849733480394e-06,
+      "loss": 0.0661,
+      "step": 361
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.0290681996060605e-06,
+      "loss": 0.0362,
+      "step": 362
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.974427779587004e-06,
+      "loss": 0.0815,
+      "step": 363
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.919930415164033e-06,
+      "loss": 0.0205,
+      "step": 364
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.865578042994227e-06,
+      "loss": 0.0065,
+      "step": 365
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.8113725945821245e-06,
+      "loss": 0.2377,
+      "step": 366
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.757315996211066e-06,
+      "loss": 0.0673,
+      "step": 367
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.703410168874768e-06,
+      "loss": 0.1033,
+      "step": 368
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.649657028209024e-06,
+      "loss": 0.1259,
+      "step": 369
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.5960584844236565e-06,
+      "loss": 0.0052,
+      "step": 370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.542616442234618e-06,
+      "loss": 0.1048,
+      "step": 371
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.48933280079631e-06,
+      "loss": 0.3342,
+      "step": 372
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.436209453634087e-06,
+      "loss": 0.0725,
+      "step": 373
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3832482885769855e-06,
+      "loss": 0.1597,
+      "step": 374
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.330451187690614e-06,
+      "loss": 0.2186,
+      "step": 375
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.277820027210279e-06,
+      "loss": 0.0521,
+      "step": 376
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.225356677474309e-06,
+      "loss": 0.0426,
+      "step": 377
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.1730630028576055e-06,
+      "loss": 0.1171,
+      "step": 378
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.120940861705357e-06,
+      "loss": 0.0551,
+      "step": 379
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.068992106267021e-06,
+      "loss": 0.1238,
+      "step": 380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 5.017218582630507e-06,
+      "loss": 0.4425,
+      "step": 381
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.965622130656551e-06,
+      "loss": 0.1591,
+      "step": 382
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.914204583913349e-06,
+      "loss": 0.0568,
+      "step": 383
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.862967769611389e-06,
+      "loss": 0.0159,
+      "step": 384
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.8119135085385375e-06,
+      "loss": 0.055,
+      "step": 385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.7610436149953e-06,
+      "loss": 0.0356,
+      "step": 386
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.710359896730379e-06,
+      "loss": 0.0969,
+      "step": 387
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.659864154876411e-06,
+      "loss": 0.1161,
+      "step": 388
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.609558183885979e-06,
+      "loss": 0.0437,
+      "step": 389
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.559443771467833e-06,
+      "loss": 0.1526,
+      "step": 390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.509522698523352e-06,
+      "loss": 0.0183,
+      "step": 391
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.4597967390832745e-06,
+      "loss": 0.073,
+      "step": 392
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.4102676602446375e-06,
+      "loss": 0.0411,
+      "step": 393
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.360937222108002e-06,
+      "loss": 0.0524,
+      "step": 394
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3118071777148865e-06,
+      "loss": 0.1156,
+      "step": 395
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.262879272985468e-06,
+      "loss": 0.0311,
+      "step": 396
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.21415524665655e-06,
+      "loss": 0.1253,
+      "step": 397
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.165636830219776e-06,
+      "loss": 0.0589,
+      "step": 398
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.117325747860077e-06,
+      "loss": 0.0248,
+      "step": 399
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.069223716394419e-06,
+      "loss": 0.0164,
+      "step": 400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.021332445210785e-06,
+      "loss": 0.1801,
+      "step": 401
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.973653636207437e-06,
+      "loss": 0.107,
+      "step": 402
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.9261889837324245e-06,
+      "loss": 0.0477,
+      "step": 403
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.878940174523371e-06,
+      "loss": 0.0214,
+      "step": 404
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.8319088876475595e-06,
+      "loss": 0.1071,
+      "step": 405
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.785096794442229e-06,
+      "loss": 0.071,
+      "step": 406
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7385055584552e-06,
+      "loss": 0.0623,
+      "step": 407
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.6921368353857524e-06,
+      "loss": 0.0534,
+      "step": 408
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.645992273025797e-06,
+      "loss": 0.1143,
+      "step": 409
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.6000735112012984e-06,
+      "loss": 0.1056,
+      "step": 410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5543821817140313e-06,
+      "loss": 0.0537,
+      "step": 411
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5089199082835436e-06,
+      "loss": 0.0065,
+      "step": 412
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.463688306489511e-06,
+      "loss": 0.0995,
+      "step": 413
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.418688983714291e-06,
+      "loss": 0.0818,
+      "step": 414
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.373923539085805e-06,
+      "loss": 0.0481,
+      "step": 415
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.329393563420713e-06,
+      "loss": 0.1379,
+      "step": 416
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.285100639167883e-06,
+      "loss": 0.1759,
+      "step": 417
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2410463403521653e-06,
+      "loss": 0.0599,
+      "step": 418
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1972322325184347e-06,
+      "loss": 0.0898,
+      "step": 419
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.1536598726759747e-06,
+      "loss": 0.0079,
+      "step": 420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.110330809243134e-06,
+      "loss": 0.0185,
+      "step": 421
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0672465819923215e-06,
+      "loss": 0.0792,
+      "step": 422
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0244087219952565e-06,
+      "loss": 0.1059,
+      "step": 423
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.981818751568586e-06,
+      "loss": 0.044,
+      "step": 424
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.939478184219777e-06,
+      "loss": 0.0766,
+      "step": 425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8973885245933287e-06,
+      "loss": 0.1558,
+      "step": 426
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.855551268417305e-06,
+      "loss": 0.0052,
+      "step": 427
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.813967902450179e-06,
+      "loss": 0.0747,
+      "step": 428
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7726399044280107e-06,
+      "loss": 0.0868,
+      "step": 429
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7315687430119097e-06,
+      "loss": 0.047,
+      "step": 430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6907558777358756e-06,
+      "loss": 0.0721,
+      "step": 431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.650202758954886e-06,
+      "loss": 0.128,
+      "step": 432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6099108277934105e-06,
+      "loss": 0.08,
+      "step": 433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5698815160941494e-06,
+      "loss": 0.0901,
+      "step": 434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5301162463671845e-06,
+      "loss": 0.0965,
+      "step": 435
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4906164317394067e-06,
+      "loss": 0.062,
+      "step": 436
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.451383475904304e-06,
+      "loss": 0.0634,
+      "step": 437
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4124187730720916e-06,
+      "loss": 0.1525,
+      "step": 438
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3737237079201437e-06,
+      "loss": 0.1071,
+      "step": 439
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3352996555438036e-06,
+      "loss": 0.0409,
+      "step": 440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.297147981407509e-06,
+      "loss": 0.1753,
+      "step": 441
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.2592700412962775e-06,
+      "loss": 0.175,
+      "step": 442
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.2216671812675118e-06,
+      "loss": 0.0348,
+      "step": 443
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.184340737603178e-06,
+      "loss": 0.105,
+      "step": 444
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1472920367623094e-06,
+      "loss": 0.0477,
+      "step": 445
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1105223953338805e-06,
+      "loss": 0.0176,
+      "step": 446
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0740331199900053e-06,
+      "loss": 0.6195,
+      "step": 447
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0378255074395094e-06,
+      "loss": 0.0913,
+      "step": 448
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.001900844381857e-06,
+      "loss": 0.0386,
+      "step": 449
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9662604074614044e-06,
+      "loss": 0.1309,
+      "step": 450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9309054632220645e-06,
+      "loss": 0.0218,
+      "step": 451
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.895837268062256e-06,
+      "loss": 0.0185,
+      "step": 452
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8610570681903018e-06,
+      "loss": 0.3416,
+      "step": 453
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8265660995801004e-06,
+      "loss": 0.2817,
+      "step": 454
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7923655879272395e-06,
+      "loss": 0.0182,
+      "step": 455
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7584567486054039e-06,
+      "loss": 0.0665,
+      "step": 456
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7248407866232175e-06,
+      "loss": 0.0403,
+      "step": 457
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6915188965814034e-06,
+      "loss": 0.017,
+      "step": 458
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6915188965814034e-06,
+      "loss": 0.3175,
+      "step": 459
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6584922626303325e-06,
+      "loss": 0.0474,
+      "step": 460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6257620584279454e-06,
+      "loss": 0.0881,
+      "step": 461
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5933294470980443e-06,
+      "loss": 0.0475,
+      "step": 462
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5611955811889645e-06,
+      "loss": 0.0473,
+      "step": 463
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5293616026326053e-06,
+      "loss": 0.0143,
+      "step": 464
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4978286427038602e-06,
+      "loss": 0.1228,
+      "step": 465
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4665978219804056e-06,
+      "loss": 0.2635,
+      "step": 466
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.435670250302892e-06,
+      "loss": 0.0668,
+      "step": 467
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.405047026735491e-06,
+      "loss": 0.082,
+      "step": 468
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.3747292395268407e-06,
+      "loss": 0.085,
+      "step": 469
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.344717966071385e-06,
+      "loss": 0.1178,
+      "step": 470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.3150142728710669e-06,
+      "loss": 0.0633,
+      "step": 471
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2856192154974488e-06,
+      "loss": 0.0229,
+      "step": 472
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2565338385541792e-06,
+      "loss": 0.0356,
+      "step": 473
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2277591756398933e-06,
+      "loss": 0.1599,
+      "step": 474
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1992962493114645e-06,
+      "loss": 0.0168,
+      "step": 475
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.171146071047683e-06,
+      "loss": 0.0626,
+      "step": 476
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1433096412132838e-06,
+      "loss": 0.1343,
+      "step": 477
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1157879490234346e-06,
+      "loss": 0.0529,
+      "step": 478
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.088581972508549e-06,
+      "loss": 0.0556,
+      "step": 479
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0616926784795511e-06,
+      "loss": 0.0903,
+      "step": 480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.035121022493506e-06,
+      "loss": 0.0993,
+      "step": 481
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0088679488196695e-06,
+      "loss": 0.0673,
+      "step": 482
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.829343904059342e-07,
+      "loss": 0.018,
+      "step": 483
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.573212688456635e-07,
+      "loss": 0.1005,
+      "step": 484
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.320294943449537e-07,
+      "loss": 0.0859,
+      "step": 485
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.070599656902801e-07,
+      "loss": 0.0361,
+      "step": 486
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.824135702165693e-07,
+      "loss": 0.0256,
+      "step": 487
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.580911837756467e-07,
+      "loss": 0.0652,
+      "step": 488
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.340936707051273e-07,
+      "loss": 0.103,
+      "step": 489
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.10421883797694e-07,
+      "loss": 0.0589,
+      "step": 490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.87076664270795e-07,
+      "loss": 0.1919,
+      "step": 491
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.6405884173675e-07,
+      "loss": 0.1313,
+      "step": 492
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.413692341732582e-07,
+      "loss": 0.0657,
+      "step": 493
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.190086478943459e-07,
+      "loss": 0.1785,
+      "step": 494
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.969778775217007e-07,
+      "loss": 0.1866,
+      "step": 495
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.752777059564431e-07,
+      "loss": 0.0295,
+      "step": 496
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.539089043512914e-07,
+      "loss": 0.0316,
+      "step": 497
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.328722320831737e-07,
+      "loss": 0.0702,
+      "step": 498
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.121684367262271e-07,
+      "loss": 0.0271,
+      "step": 499
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.917982540252442e-07,
+      "loss": 0.0398,
+      "step": 500
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 555,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "total_flos": 1080325140480.0,
+  "train_batch_size": 10,
+  "trial_name": null,
+  "trial_params": null
+}

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19850de0da1b4eac2d894ea196e517fb53755d5462a1803598ffe2ca47c1ad86
+size 6968

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/zero_to_fp32.py

@@ -0,0 +1,587 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel()
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file, tag=args.tag)

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/config.json

@@ -0,0 +1,70 @@
+{
+  "_name_or_path": "liuhaotian/llava-v1.6-mistral-7b",
+  "architectures": [
+    "LlavaMistralForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "freeze_mm_mlp_adapter": false,
+  "freeze_mm_vision_resampler": false,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "image_aspect_ratio": "anyres",
+  "image_crop_resolution": 224,
+  "image_grid_pinpoints": [
+    [
+      336,
+      672
+    ],
+    [
+      672,
+      336
+    ],
+    [
+      672,
+      672
+    ],
+    [
+      1008,
+      336
+    ],
+    [
+      336,
+      1008
+    ]
+  ],
+  "image_split_resolution": 224,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 32768,
+  "mm_hidden_size": 1024,
+  "mm_patch_merge_type": "spatial_unpad",
+  "mm_projector_lr": 2e-05,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "openai/clip-vit-large-patch14-336",
+  "mm_vision_tower_lr": 2e-06,
+  "model_type": "llava_mistral",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "rms_norm_eps": 1e-05,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "tokenizer_model_max_length": 4096,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.37.2",
+  "tune_mm_mlp_adapter": false,
+  "tune_mm_vision_resampler": false,
+  "unfreeze_mm_vision_tower": true,
+  "use_cache": true,
+  "use_mm_proj": true,
+  "vocab_size": 32000
+}

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/non_lora_trainables.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fde24a5cdabb49bef91a9dd1ee36c3b2ed72791efc564f6476124852a334852
+size 41961648

CheckGuard Models/wholeimage/bank/llava-lora-mistral-r128a256-10BS-model/trainer_state.json

@@ -0,0 +1,3360 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 555,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.142857142857143e-07,
+      "loss": 0.4237,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4285714285714286e-06,
+      "loss": 0.3368,
+      "step": 2
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1428571428571427e-06,
+      "loss": 0.214,
+      "step": 3
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8571428571428573e-06,
+      "loss": 0.396,
+      "step": 4
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5714285714285718e-06,
+      "loss": 0.305,
+      "step": 5
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2857142857142855e-06,
+      "loss": 0.4049,
+      "step": 6
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5e-06,
+      "loss": 0.108,
+      "step": 7
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.7142857142857145e-06,
+      "loss": 0.2286,
+      "step": 8
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.4285714285714295e-06,
+      "loss": 0.1443,
+      "step": 9
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.1428571428571436e-06,
+      "loss": 0.2252,
+      "step": 10
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.857142857142858e-06,
+      "loss": 0.0747,
+      "step": 11
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 0.1084,
+      "step": 12
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.285714285714288e-06,
+      "loss": 0.2115,
+      "step": 13
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-05,
+      "loss": 0.4742,
+      "step": 14
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0714285714285714e-05,
+      "loss": 0.083,
+      "step": 15
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 0.3392,
+      "step": 16
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.2142857142857142e-05,
+      "loss": 0.065,
+      "step": 17
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.2857142857142859e-05,
+      "loss": 0.1711,
+      "step": 18
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.3571428571428574e-05,
+      "loss": 0.0539,
+      "step": 19
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4285714285714287e-05,
+      "loss": 0.0701,
+      "step": 20
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.0836,
+      "step": 21
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.5714285714285715e-05,
+      "loss": 0.1891,
+      "step": 22
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.642857142857143e-05,
+      "loss": 0.0422,
+      "step": 23
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.7142857142857142e-05,
+      "loss": 0.2094,
+      "step": 24
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.785714285714286e-05,
+      "loss": 0.139,
+      "step": 25
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8571428571428575e-05,
+      "loss": 0.2214,
+      "step": 26
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.928571428571429e-05,
+      "loss": 0.1084,
+      "step": 27
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2e-05,
+      "loss": 0.0898,
+      "step": 28
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9999822316445652e-05,
+      "loss": 0.0359,
+      "step": 29
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9999289272096886e-05,
+      "loss": 0.2648,
+      "step": 30
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9998400885896355e-05,
+      "loss": 0.4007,
+      "step": 31
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9997157189414373e-05,
+      "loss": 0.235,
+      "step": 32
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999555822684783e-05,
+      "loss": 0.0273,
+      "step": 33
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999360405501859e-05,
+      "loss": 0.0267,
+      "step": 34
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99912947433715e-05,
+      "loss": 0.2619,
+      "step": 35
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9988630373971896e-05,
+      "loss": 0.4101,
+      "step": 36
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9985611041502704e-05,
+      "loss": 0.1302,
+      "step": 37
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9982236853261067e-05,
+      "loss": 0.118,
+      "step": 38
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9978507929154534e-05,
+      "loss": 0.0933,
+      "step": 39
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997442440169681e-05,
+      "loss": 0.0104,
+      "step": 40
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9969986416003026e-05,
+      "loss": 0.1061,
+      "step": 41
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965194129784597e-05,
+      "loss": 0.1575,
+      "step": 42
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996004771334361e-05,
+      "loss": 0.1969,
+      "step": 43
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996004771334361e-05,
+      "loss": 0.0492,
+      "step": 44
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9954547349566783e-05,
+      "loss": 0.3012,
+      "step": 45
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.994869323391895e-05,
+      "loss": 0.2185,
+      "step": 46
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.994248557443613e-05,
+      "loss": 0.1729,
+      "step": 47
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.993592459171812e-05,
+      "loss": 0.0354,
+      "step": 48
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9929010518920667e-05,
+      "loss": 0.3939,
+      "step": 49
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.992174360174717e-05,
+      "loss": 0.0505,
+      "step": 50
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9914124098439976e-05,
+      "loss": 0.0777,
+      "step": 51
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9914124098439976e-05,
+      "loss": 0.6651,
+      "step": 52
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9906152279771162e-05,
+      "loss": 0.15,
+      "step": 53
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9897828429032946e-05,
+      "loss": 0.1416,
+      "step": 54
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9889152842027607e-05,
+      "loss": 0.1195,
+      "step": 55
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9880125827056967e-05,
+      "loss": 0.0787,
+      "step": 56
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.987074770491145e-05,
+      "loss": 0.0681,
+      "step": 57
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.986101880885867e-05,
+      "loss": 0.1337,
+      "step": 58
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9850939484631598e-05,
+      "loss": 0.0961,
+      "step": 59
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.984051009041626e-05,
+      "loss": 0.116,
+      "step": 60
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.982973099683902e-05,
+      "loss": 0.3853,
+      "step": 61
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9818602586953414e-05,
+      "loss": 0.0875,
+      "step": 62
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9807125256226532e-05,
+      "loss": 0.3216,
+      "step": 63
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9795299412524948e-05,
+      "loss": 0.0752,
+      "step": 64
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9783125476100254e-05,
+      "loss": 0.1461,
+      "step": 65
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9770603879574108e-05,
+      "loss": 0.075,
+      "step": 66
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.975773506792287e-05,
+      "loss": 0.0685,
+      "step": 67
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.974451949846177e-05,
+      "loss": 0.0555,
+      "step": 68
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.973095764082869e-05,
+      "loss": 0.0171,
+      "step": 69
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9717049976967437e-05,
+      "loss": 0.0247,
+      "step": 70
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9702797001110642e-05,
+      "loss": 0.0839,
+      "step": 71
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9688199219762183e-05,
+      "loss": 0.4163,
+      "step": 72
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.96732571516792e-05,
+      "loss": 0.1461,
+      "step": 73
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9657971327853644e-05,
+      "loss": 0.1457,
+      "step": 74
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.964234229149342e-05,
+      "loss": 0.0482,
+      "step": 75
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.962637059800307e-05,
+      "loss": 0.0802,
+      "step": 76
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9610056814964053e-05,
+      "loss": 0.0697,
+      "step": 77
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.959340152211455e-05,
+      "loss": 0.0614,
+      "step": 78
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.95764053113289e-05,
+      "loss": 0.1004,
+      "step": 79
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9559068786596526e-05,
+      "loss": 0.0286,
+      "step": 80
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.954139256400049e-05,
+      "loss": 0.1162,
+      "step": 81
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.952337727169561e-05,
+      "loss": 0.0731,
+      "step": 82
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.950502354988612e-05,
+      "loss": 0.0286,
+      "step": 83
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.948633205080292e-05,
+      "loss": 0.2425,
+      "step": 84
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9467303438680414e-05,
+      "loss": 0.0505,
+      "step": 85
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.944793838973289e-05,
+      "loss": 0.0922,
+      "step": 86
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9428237592130487e-05,
+      "loss": 0.2949,
+      "step": 87
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.940820174597476e-05,
+      "loss": 0.2807,
+      "step": 88
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9387831563273775e-05,
+      "loss": 0.2377,
+      "step": 89
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9367127767916828e-05,
+      "loss": 0.2558,
+      "step": 90
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9346091095648712e-05,
+      "loss": 0.0871,
+      "step": 91
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.932472229404356e-05,
+      "loss": 0.2204,
+      "step": 92
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9303022122478303e-05,
+      "loss": 0.1174,
+      "step": 93
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9280991352105656e-05,
+      "loss": 0.2181,
+      "step": 94
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.925863076582674e-05,
+      "loss": 0.1251,
+      "step": 95
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9235941158263253e-05,
+      "loss": 0.2251,
+      "step": 96
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9212923335729206e-05,
+      "loss": 0.1236,
+      "step": 97
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.918957811620231e-05,
+      "loss": 0.0901,
+      "step": 98
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9165906329294875e-05,
+      "loss": 0.1002,
+      "step": 99
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9141908816224356e-05,
+      "loss": 0.4397,
+      "step": 100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9117586429783433e-05,
+      "loss": 0.1141,
+      "step": 101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.909294003430972e-05,
+      "loss": 0.1842,
+      "step": 102
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.906797050565505e-05,
+      "loss": 0.0985,
+      "step": 103
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9042678731154337e-05,
+      "loss": 0.1533,
+      "step": 104
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.901706560959407e-05,
+      "loss": 0.145,
+      "step": 105
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8991132051180332e-05,
+      "loss": 0.1693,
+      "step": 106
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8964878977506496e-05,
+      "loss": 0.2012,
+      "step": 107
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8938307321520453e-05,
+      "loss": 0.1286,
+      "step": 108
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8911418027491453e-05,
+      "loss": 0.1396,
+      "step": 109
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8884212050976568e-05,
+      "loss": 0.0291,
+      "step": 110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.885669035878672e-05,
+      "loss": 0.0317,
+      "step": 111
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.882885392895232e-05,
+      "loss": 0.1143,
+      "step": 112
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8800703750688536e-05,
+      "loss": 0.126,
+      "step": 113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.877224082436011e-05,
+      "loss": 0.2017,
+      "step": 114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8743466161445823e-05,
+      "loss": 0.0735,
+      "step": 115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8714380784502553e-05,
+      "loss": 0.0527,
+      "step": 116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8684985727128936e-05,
+      "loss": 0.1112,
+      "step": 117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8655282033928618e-05,
+      "loss": 0.3129,
+      "step": 118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8625270760473164e-05,
+      "loss": 0.2827,
+      "step": 119
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8594952973264512e-05,
+      "loss": 0.5608,
+      "step": 120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856432974969711e-05,
+      "loss": 0.1465,
+      "step": 121
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8533402178019596e-05,
+      "loss": 0.1322,
+      "step": 122
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8502171357296144e-05,
+      "loss": 0.0912,
+      "step": 123
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8470638397367397e-05,
+      "loss": 0.0419,
+      "step": 124
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8438804418811038e-05,
+      "loss": 0.0369,
+      "step": 125
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8406670552901958e-05,
+      "loss": 0.0529,
+      "step": 126
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.837423794157206e-05,
+      "loss": 0.1472,
+      "step": 127
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.834150773736967e-05,
+      "loss": 0.0425,
+      "step": 128
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8308481103418597e-05,
+      "loss": 0.1634,
+      "step": 129
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8275159213376783e-05,
+      "loss": 0.0485,
+      "step": 130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.82415432513946e-05,
+      "loss": 0.0313,
+      "step": 131
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8207634412072765e-05,
+      "loss": 0.1792,
+      "step": 132
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.81734339004199e-05,
+      "loss": 0.1184,
+      "step": 133
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8138942931809702e-05,
+      "loss": 0.2756,
+      "step": 134
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8104162731937746e-05,
+      "loss": 0.0635,
+      "step": 135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8069094536777938e-05,
+      "loss": 0.0158,
+      "step": 136
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8033739592538598e-05,
+      "loss": 0.2732,
+      "step": 137
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7998099155618147e-05,
+      "loss": 0.1428,
+      "step": 138
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7962174492560492e-05,
+      "loss": 0.0777,
+      "step": 139
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7925966880009998e-05,
+      "loss": 0.1644,
+      "step": 140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7889477604666124e-05,
+      "loss": 0.0999,
+      "step": 141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.785270796323769e-05,
+      "loss": 0.0446,
+      "step": 142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7815659262396825e-05,
+      "loss": 0.0647,
+      "step": 143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7778332818732492e-05,
+      "loss": 0.0521,
+      "step": 144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7740729958703725e-05,
+      "loss": 0.2041,
+      "step": 145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7702852018592493e-05,
+      "loss": 0.0149,
+      "step": 146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7664700344456198e-05,
+      "loss": 0.0502,
+      "step": 147
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.762627629207986e-05,
+      "loss": 0.2027,
+      "step": 148
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.758758122692791e-05,
+      "loss": 0.0187,
+      "step": 149
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7548616524095697e-05,
+      "loss": 0.1248,
+      "step": 150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7509383568260597e-05,
+      "loss": 0.0859,
+      "step": 151
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7469883753632817e-05,
+      "loss": 0.0822,
+      "step": 152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.743011848390585e-05,
+      "loss": 0.2445,
+      "step": 153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7390089172206594e-05,
+      "loss": 0.2662,
+      "step": 154
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7349797241045115e-05,
+      "loss": 0.0984,
+      "step": 155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.730924412226413e-05,
+      "loss": 0.0317,
+      "step": 156
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.726843125698809e-05,
+      "loss": 0.1129,
+      "step": 157
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7227360095571992e-05,
+      "loss": 0.1882,
+      "step": 158
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7186032097549822e-05,
+      "loss": 0.1099,
+      "step": 159
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7144448731582698e-05,
+      "loss": 0.3506,
+      "step": 160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7102611475406676e-05,
+      "loss": 0.0936,
+      "step": 161
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7060521815780225e-05,
+      "loss": 0.104,
+      "step": 162
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7018181248431416e-05,
+      "loss": 0.168,
+      "step": 163
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6975591278004747e-05,
+      "loss": 0.2726,
+      "step": 164
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6932753418007683e-05,
+      "loss": 0.0564,
+      "step": 165
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.688966919075687e-05,
+      "loss": 0.2981,
+      "step": 166
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.684634012732403e-05,
+      "loss": 0.0602,
+      "step": 167
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.680276776748157e-05,
+      "loss": 0.0364,
+      "step": 168
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6758953659647838e-05,
+      "loss": 0.096,
+      "step": 169
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6714899360832118e-05,
+      "loss": 0.2139,
+      "step": 170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.667060643657929e-05,
+      "loss": 0.1666,
+      "step": 171
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.66260764609142e-05,
+      "loss": 0.0486,
+      "step": 172
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.658131101628571e-05,
+      "loss": 0.055,
+      "step": 173
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.653631169351049e-05,
+      "loss": 0.0953,
+      "step": 174
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6491080091716457e-05,
+      "loss": 0.1824,
+      "step": 175
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6445617818285974e-05,
+      "loss": 0.0226,
+      "step": 176
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6399926488798702e-05,
+      "loss": 0.0388,
+      "step": 177
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6354007726974205e-05,
+      "loss": 0.1149,
+      "step": 178
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.630786316461425e-05,
+      "loss": 0.1428,
+      "step": 179
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6261494441544805e-05,
+      "loss": 0.0445,
+      "step": 180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6214903205557774e-05,
+      "loss": 0.0612,
+      "step": 181
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6168091112352443e-05,
+      "loss": 0.0826,
+      "step": 182
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.612105982547663e-05,
+      "loss": 0.0376,
+      "step": 183
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.607381101626758e-05,
+      "loss": 0.1441,
+      "step": 184
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6026346363792565e-05,
+      "loss": 0.1089,
+      "step": 185
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5978667554789216e-05,
+      "loss": 0.0845,
+      "step": 186
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5930776283605585e-05,
+      "loss": 0.0835,
+      "step": 187
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5882674252139928e-05,
+      "loss": 0.0762,
+      "step": 188
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5834363169780227e-05,
+      "loss": 0.067,
+      "step": 189
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.578584475334345e-05,
+      "loss": 0.0327,
+      "step": 190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5737120727014535e-05,
+      "loss": 0.0254,
+      "step": 191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5688192822285116e-05,
+      "loss": 0.028,
+      "step": 192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5639062777892e-05,
+      "loss": 0.1708,
+      "step": 193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5589732339755362e-05,
+      "loss": 0.0542,
+      "step": 194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5540203260916728e-05,
+      "loss": 0.2358,
+      "step": 195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5490477301476648e-05,
+      "loss": 0.1471,
+      "step": 196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5440556228532168e-05,
+      "loss": 0.0414,
+      "step": 197
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5390441816114022e-05,
+      "loss": 0.0754,
+      "step": 198
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.534013584512359e-05,
+      "loss": 0.105,
+      "step": 199
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5289640103269626e-05,
+      "loss": 0.2052,
+      "step": 200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5238956385004703e-05,
+      "loss": 0.2482,
+      "step": 201
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5188086491461467e-05,
+      "loss": 0.0967,
+      "step": 202
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5137032230388613e-05,
+      "loss": 0.1314,
+      "step": 203
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5085795416086655e-05,
+      "loss": 0.2313,
+      "step": 204
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5034377869343453e-05,
+      "loss": 0.1304,
+      "step": 205
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4982781417369496e-05,
+      "loss": 0.2304,
+      "step": 206
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4931007893732981e-05,
+      "loss": 0.0508,
+      "step": 207
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4879059138294647e-05,
+      "loss": 0.1389,
+      "step": 208
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4826936997142399e-05,
+      "loss": 0.2129,
+      "step": 209
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4774643322525691e-05,
+      "loss": 0.0201,
+      "step": 210
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4722179972789725e-05,
+      "loss": 0.1064,
+      "step": 211
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.466954881230939e-05,
+      "loss": 0.0459,
+      "step": 212
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4616751711423016e-05,
+      "loss": 0.2229,
+      "step": 213
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4563790546365914e-05,
+      "loss": 0.1464,
+      "step": 214
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4510667199203697e-05,
+      "loss": 0.0558,
+      "step": 215
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4457383557765385e-05,
+      "loss": 0.0214,
+      "step": 216
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4403941515576344e-05,
+      "loss": 0.1551,
+      "step": 217
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4350342971790979e-05,
+      "loss": 0.2093,
+      "step": 218
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4296589831125234e-05,
+      "loss": 0.0453,
+      "step": 219
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4242684003788934e-05,
+      "loss": 0.0317,
+      "step": 220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.418862740541788e-05,
+      "loss": 0.1334,
+      "step": 221
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4134421957005775e-05,
+      "loss": 0.0185,
+      "step": 222
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4080069584835971e-05,
+      "loss": 0.087,
+      "step": 223
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4025572220412998e-05,
+      "loss": 0.1747,
+      "step": 224
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3970931800393943e-05,
+      "loss": 0.1168,
+      "step": 225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.391615026651961e-05,
+      "loss": 0.5095,
+      "step": 226
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3861229565545532e-05,
+      "loss": 0.1157,
+      "step": 227
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3806171649172782e-05,
+      "loss": 0.1201,
+      "step": 228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3750978473978611e-05,
+      "loss": 0.2232,
+      "step": 229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3695652001346928e-05,
+      "loss": 0.1718,
+      "step": 230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.36401941973986e-05,
+      "loss": 0.0509,
+      "step": 231
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3584607032921566e-05,
+      "loss": 0.0333,
+      "step": 232
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3528892483300821e-05,
+      "loss": 0.1811,
+      "step": 233
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3473052528448203e-05,
+      "loss": 0.1771,
+      "step": 234
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3417089152732049e-05,
+      "loss": 0.1098,
+      "step": 235
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3361004344906652e-05,
+      "loss": 0.0566,
+      "step": 236
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.330480009804162e-05,
+      "loss": 0.2864,
+      "step": 237
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3248478409451017e-05,
+      "loss": 0.0166,
+      "step": 238
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3192041280622409e-05,
+      "loss": 0.2239,
+      "step": 239
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3135490717145726e-05,
+      "loss": 0.2247,
+      "step": 240
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3078828728641994e-05,
+      "loss": 0.1758,
+      "step": 241
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.3022057328691915e-05,
+      "loss": 0.0618,
+      "step": 242
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2965178534764311e-05,
+      "loss": 0.1204,
+      "step": 243
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2908194368144437e-05,
+      "loss": 0.0233,
+      "step": 244
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.285110685386215e-05,
+      "loss": 0.0387,
+      "step": 245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2793918020619937e-05,
+      "loss": 0.0791,
+      "step": 246
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2736629900720832e-05,
+      "loss": 0.0106,
+      "step": 247
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2679244529996182e-05,
+      "loss": 0.042,
+      "step": 248
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.262176394773332e-05,
+      "loss": 0.0725,
+      "step": 249
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.256419019660308e-05,
+      "loss": 0.0834,
+      "step": 250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2506525322587207e-05,
+      "loss": 0.0432,
+      "step": 251
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2448771374905655e-05,
+      "loss": 0.177,
+      "step": 252
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2390930405943766e-05,
+      "loss": 0.0887,
+      "step": 253
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.233300447117933e-05,
+      "loss": 0.0152,
+      "step": 254
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2274995629109545e-05,
+      "loss": 0.0317,
+      "step": 255
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2216905941177854e-05,
+      "loss": 0.0268,
+      "step": 256
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.215873747170071e-05,
+      "loss": 0.1685,
+      "step": 257
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2100492287794186e-05,
+      "loss": 0.1403,
+      "step": 258
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.2042172459300546e-05,
+      "loss": 0.0443,
+      "step": 259
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.198378005871467e-05,
+      "loss": 0.3589,
+      "step": 260
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.192531716111042e-05,
+      "loss": 0.0427,
+      "step": 261
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1866785844066884e-05,
+      "loss": 0.1103,
+      "step": 262
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1808188187594549e-05,
+      "loss": 0.2563,
+      "step": 263
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1749526274061394e-05,
+      "loss": 0.1494,
+      "step": 264
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1690802188118878e-05,
+      "loss": 0.1105,
+      "step": 265
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1632018016627859e-05,
+      "loss": 0.082,
+      "step": 266
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1573175848584455e-05,
+      "loss": 0.3555,
+      "step": 267
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1514277775045768e-05,
+      "loss": 0.0603,
+      "step": 268
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1455325889055616e-05,
+      "loss": 0.2883,
+      "step": 269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1396322285570119e-05,
+      "loss": 0.054,
+      "step": 270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1337269061383278e-05,
+      "loss": 0.0668,
+      "step": 271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1278168315052445e-05,
+      "loss": 0.1454,
+      "step": 272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1219022146823762e-05,
+      "loss": 0.0619,
+      "step": 273
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1159832658557498e-05,
+      "loss": 0.0449,
+      "step": 274
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1100601953653393e-05,
+      "loss": 0.0684,
+      "step": 275
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1041332136975874e-05,
+      "loss": 0.0273,
+      "step": 276
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0982025314779287e-05,
+      "loss": 0.2375,
+      "step": 277
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.092268359463302e-05,
+      "loss": 0.0353,
+      "step": 278
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.086330908534663e-05,
+      "loss": 0.1224,
+      "step": 279
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0803903896894877e-05,
+      "loss": 0.1297,
+      "step": 280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0744470140342775e-05,
+      "loss": 0.4464,
+      "step": 281
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0685009927770542e-05,
+      "loss": 0.103,
+      "step": 282
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0625525372198564e-05,
+      "loss": 0.0881,
+      "step": 283
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.056601858751229e-05,
+      "loss": 0.075,
+      "step": 284
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0506491688387128e-05,
+      "loss": 0.0677,
+      "step": 285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0446946790213275e-05,
+      "loss": 0.2301,
+      "step": 286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0387386009020569e-05,
+      "loss": 0.0737,
+      "step": 287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.032781146140326e-05,
+      "loss": 0.1262,
+      "step": 288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0268225264444829e-05,
+      "loss": 0.0252,
+      "step": 289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0208629535642726e-05,
+      "loss": 0.0192,
+      "step": 290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0149026392833137e-05,
+      "loss": 0.257,
+      "step": 291
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0089417954115715e-05,
+      "loss": 0.1876,
+      "step": 292
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.002980633777831e-05,
+      "loss": 0.0341,
+      "step": 293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.970193662221694e-06,
+      "loss": 0.232,
+      "step": 294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.910582045884292e-06,
+      "loss": 0.1429,
+      "step": 295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.850973607166865e-06,
+      "loss": 0.2432,
+      "step": 296
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.791370464357279e-06,
+      "loss": 0.0288,
+      "step": 297
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.731774735555174e-06,
+      "loss": 0.2272,
+      "step": 298
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.672188538596746e-06,
+      "loss": 0.1102,
+      "step": 299
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.612613990979436e-06,
+      "loss": 0.0529,
+      "step": 300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.553053209786725e-06,
+      "loss": 0.1721,
+      "step": 301
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.493508311612874e-06,
+      "loss": 0.0046,
+      "step": 302
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.433981412487711e-06,
+      "loss": 0.043,
+      "step": 303
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.374474627801439e-06,
+      "loss": 0.0589,
+      "step": 304
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.314990072229461e-06,
+      "loss": 0.0114,
+      "step": 305
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.25552985965723e-06,
+      "loss": 0.1645,
+      "step": 306
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.196096103105127e-06,
+      "loss": 0.2002,
+      "step": 307
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.136690914653377e-06,
+      "loss": 0.057,
+      "step": 308
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.07731640536698e-06,
+      "loss": 0.1744,
+      "step": 309
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.017974685220716e-06,
+      "loss": 0.0343,
+      "step": 310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.958667863024127e-06,
+      "loss": 0.0405,
+      "step": 311
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.899398046346608e-06,
+      "loss": 0.2055,
+      "step": 312
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.840167341442505e-06,
+      "loss": 0.0673,
+      "step": 313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.78097785317624e-06,
+      "loss": 0.0291,
+      "step": 314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.721831684947557e-06,
+      "loss": 0.2443,
+      "step": 315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.662730938616724e-06,
+      "loss": 0.058,
+      "step": 316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.603677714429888e-06,
+      "loss": 0.2347,
+      "step": 317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.54467411094439e-06,
+      "loss": 0.0307,
+      "step": 318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.485722224954237e-06,
+      "loss": 0.0094,
+      "step": 319
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.426824151415548e-06,
+      "loss": 0.0724,
+      "step": 320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.367981983372143e-06,
+      "loss": 0.0816,
+      "step": 321
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.309197811881128e-06,
+      "loss": 0.0375,
+      "step": 322
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.250473725938608e-06,
+      "loss": 0.0106,
+      "step": 323
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.191811812405453e-06,
+      "loss": 0.0701,
+      "step": 324
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.133214155933118e-06,
+      "loss": 0.0134,
+      "step": 325
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.074682838889581e-06,
+      "loss": 0.1992,
+      "step": 326
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.01621994128533e-06,
+      "loss": 0.1688,
+      "step": 327
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.95782754069946e-06,
+      "loss": 0.2751,
+      "step": 328
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.899507712205818e-06,
+      "loss": 0.0192,
+      "step": 329
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.841262528299296e-06,
+      "loss": 0.0797,
+      "step": 330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.783094058822147e-06,
+      "loss": 0.0867,
+      "step": 331
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.72500437089046e-06,
+      "loss": 0.0445,
+      "step": 332
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.666995528820673e-06,
+      "loss": 0.1654,
+      "step": 333
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.609069594056234e-06,
+      "loss": 0.0168,
+      "step": 334
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.551228625094349e-06,
+      "loss": 0.0779,
+      "step": 335
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.493474677412795e-06,
+      "loss": 0.0444,
+      "step": 336
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.435809803396923e-06,
+      "loss": 0.1839,
+      "step": 337
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.37823605226668e-06,
+      "loss": 0.3834,
+      "step": 338
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.320755470003822e-06,
+      "loss": 0.0261,
+      "step": 339
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.263370099279173e-06,
+      "loss": 0.0084,
+      "step": 340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.2060819793800665e-06,
+      "loss": 0.0469,
+      "step": 341
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.148893146137852e-06,
+      "loss": 0.3605,
+      "step": 342
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.091805631855566e-06,
+      "loss": 0.0621,
+      "step": 343
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.034821465235693e-06,
+      "loss": 0.099,
+      "step": 344
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.977942671308087e-06,
+      "loss": 0.0641,
+      "step": 345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.921171271358007e-06,
+      "loss": 0.0859,
+      "step": 346
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.864509282854272e-06,
+      "loss": 0.0564,
+      "step": 347
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.8079587193775935e-06,
+      "loss": 0.0405,
+      "step": 348
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.751521590548986e-06,
+      "loss": 0.101,
+      "step": 349
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.695199901958386e-06,
+      "loss": 0.1178,
+      "step": 350
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.638995655093351e-06,
+      "loss": 0.2406,
+      "step": 351
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.582910847267957e-06,
+      "loss": 0.1846,
+      "step": 352
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.526947471551799e-06,
+      "loss": 0.1374,
+      "step": 353
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.471107516699183e-06,
+      "loss": 0.0863,
+      "step": 354
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.415392967078438e-06,
+      "loss": 0.0755,
+      "step": 355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.3598058026013995e-06,
+      "loss": 0.0732,
+      "step": 356
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.304347998653074e-06,
+      "loss": 0.0555,
+      "step": 357
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.24902152602139e-06,
+      "loss": 0.0475,
+      "step": 358
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.193828350827222e-06,
+      "loss": 0.036,
+      "step": 359
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.1387704344544684e-06,
+      "loss": 0.2679,
+      "step": 360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.083849733480394e-06,
+      "loss": 0.0661,
+      "step": 361
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.0290681996060605e-06,
+      "loss": 0.0362,
+      "step": 362
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.974427779587004e-06,
+      "loss": 0.0815,
+      "step": 363
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.919930415164033e-06,
+      "loss": 0.0205,
+      "step": 364
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.865578042994227e-06,
+      "loss": 0.0065,
+      "step": 365
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.8113725945821245e-06,
+      "loss": 0.2377,
+      "step": 366
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.757315996211066e-06,
+      "loss": 0.0673,
+      "step": 367
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.703410168874768e-06,
+      "loss": 0.1033,
+      "step": 368
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.649657028209024e-06,
+      "loss": 0.1259,
+      "step": 369
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.5960584844236565e-06,
+      "loss": 0.0052,
+      "step": 370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.542616442234618e-06,
+      "loss": 0.1048,
+      "step": 371
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.48933280079631e-06,
+      "loss": 0.3342,
+      "step": 372
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.436209453634087e-06,
+      "loss": 0.0725,
+      "step": 373
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3832482885769855e-06,
+      "loss": 0.1597,
+      "step": 374
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.330451187690614e-06,
+      "loss": 0.2186,
+      "step": 375
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.277820027210279e-06,
+      "loss": 0.0521,
+      "step": 376
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.225356677474309e-06,
+      "loss": 0.0426,
+      "step": 377
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.1730630028576055e-06,
+      "loss": 0.1171,
+      "step": 378
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.120940861705357e-06,
+      "loss": 0.0551,
+      "step": 379
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.068992106267021e-06,
+      "loss": 0.1238,
+      "step": 380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 5.017218582630507e-06,
+      "loss": 0.4425,
+      "step": 381
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.965622130656551e-06,
+      "loss": 0.1591,
+      "step": 382
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.914204583913349e-06,
+      "loss": 0.0568,
+      "step": 383
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.862967769611389e-06,
+      "loss": 0.0159,
+      "step": 384
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.8119135085385375e-06,
+      "loss": 0.055,
+      "step": 385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.7610436149953e-06,
+      "loss": 0.0356,
+      "step": 386
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.710359896730379e-06,
+      "loss": 0.0969,
+      "step": 387
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.659864154876411e-06,
+      "loss": 0.1161,
+      "step": 388
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.609558183885979e-06,
+      "loss": 0.0437,
+      "step": 389
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.559443771467833e-06,
+      "loss": 0.1526,
+      "step": 390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.509522698523352e-06,
+      "loss": 0.0183,
+      "step": 391
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.4597967390832745e-06,
+      "loss": 0.073,
+      "step": 392
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.4102676602446375e-06,
+      "loss": 0.0411,
+      "step": 393
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.360937222108002e-06,
+      "loss": 0.0524,
+      "step": 394
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3118071777148865e-06,
+      "loss": 0.1156,
+      "step": 395
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.262879272985468e-06,
+      "loss": 0.0311,
+      "step": 396
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.21415524665655e-06,
+      "loss": 0.1253,
+      "step": 397
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.165636830219776e-06,
+      "loss": 0.0589,
+      "step": 398
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.117325747860077e-06,
+      "loss": 0.0248,
+      "step": 399
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.069223716394419e-06,
+      "loss": 0.0164,
+      "step": 400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.021332445210785e-06,
+      "loss": 0.1801,
+      "step": 401
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.973653636207437e-06,
+      "loss": 0.107,
+      "step": 402
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.9261889837324245e-06,
+      "loss": 0.0477,
+      "step": 403
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.878940174523371e-06,
+      "loss": 0.0214,
+      "step": 404
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.8319088876475595e-06,
+      "loss": 0.1071,
+      "step": 405
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.785096794442229e-06,
+      "loss": 0.071,
+      "step": 406
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7385055584552e-06,
+      "loss": 0.0623,
+      "step": 407
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.6921368353857524e-06,
+      "loss": 0.0534,
+      "step": 408
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.645992273025797e-06,
+      "loss": 0.1143,
+      "step": 409
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.6000735112012984e-06,
+      "loss": 0.1056,
+      "step": 410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5543821817140313e-06,
+      "loss": 0.0537,
+      "step": 411
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5089199082835436e-06,
+      "loss": 0.0065,
+      "step": 412
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.463688306489511e-06,
+      "loss": 0.0995,
+      "step": 413
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.418688983714291e-06,
+      "loss": 0.0818,
+      "step": 414
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.373923539085805e-06,
+      "loss": 0.0481,
+      "step": 415
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.329393563420713e-06,
+      "loss": 0.1379,
+      "step": 416
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.285100639167883e-06,
+      "loss": 0.1759,
+      "step": 417
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2410463403521653e-06,
+      "loss": 0.0599,
+      "step": 418
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1972322325184347e-06,
+      "loss": 0.0898,
+      "step": 419
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.1536598726759747e-06,
+      "loss": 0.0079,
+      "step": 420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.110330809243134e-06,
+      "loss": 0.0185,
+      "step": 421
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0672465819923215e-06,
+      "loss": 0.0792,
+      "step": 422
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0244087219952565e-06,
+      "loss": 0.1059,
+      "step": 423
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.981818751568586e-06,
+      "loss": 0.044,
+      "step": 424
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.939478184219777e-06,
+      "loss": 0.0766,
+      "step": 425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8973885245933287e-06,
+      "loss": 0.1558,
+      "step": 426
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.855551268417305e-06,
+      "loss": 0.0052,
+      "step": 427
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.813967902450179e-06,
+      "loss": 0.0747,
+      "step": 428
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7726399044280107e-06,
+      "loss": 0.0868,
+      "step": 429
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7315687430119097e-06,
+      "loss": 0.047,
+      "step": 430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6907558777358756e-06,
+      "loss": 0.0721,
+      "step": 431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.650202758954886e-06,
+      "loss": 0.128,
+      "step": 432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6099108277934105e-06,
+      "loss": 0.08,
+      "step": 433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5698815160941494e-06,
+      "loss": 0.0901,
+      "step": 434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5301162463671845e-06,
+      "loss": 0.0965,
+      "step": 435
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4906164317394067e-06,
+      "loss": 0.062,
+      "step": 436
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.451383475904304e-06,
+      "loss": 0.0634,
+      "step": 437
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4124187730720916e-06,
+      "loss": 0.1525,
+      "step": 438
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3737237079201437e-06,
+      "loss": 0.1071,
+      "step": 439
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3352996555438036e-06,
+      "loss": 0.0409,
+      "step": 440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.297147981407509e-06,
+      "loss": 0.1753,
+      "step": 441
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.2592700412962775e-06,
+      "loss": 0.175,
+      "step": 442
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.2216671812675118e-06,
+      "loss": 0.0348,
+      "step": 443
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.184340737603178e-06,
+      "loss": 0.105,
+      "step": 444
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1472920367623094e-06,
+      "loss": 0.0477,
+      "step": 445
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1105223953338805e-06,
+      "loss": 0.0176,
+      "step": 446
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0740331199900053e-06,
+      "loss": 0.6195,
+      "step": 447
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0378255074395094e-06,
+      "loss": 0.0913,
+      "step": 448
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.001900844381857e-06,
+      "loss": 0.0386,
+      "step": 449
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9662604074614044e-06,
+      "loss": 0.1309,
+      "step": 450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9309054632220645e-06,
+      "loss": 0.0218,
+      "step": 451
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.895837268062256e-06,
+      "loss": 0.0185,
+      "step": 452
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8610570681903018e-06,
+      "loss": 0.3416,
+      "step": 453
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8265660995801004e-06,
+      "loss": 0.2817,
+      "step": 454
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7923655879272395e-06,
+      "loss": 0.0182,
+      "step": 455
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7584567486054039e-06,
+      "loss": 0.0665,
+      "step": 456
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7248407866232175e-06,
+      "loss": 0.0403,
+      "step": 457
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6915188965814034e-06,
+      "loss": 0.017,
+      "step": 458
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6915188965814034e-06,
+      "loss": 0.3175,
+      "step": 459
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6584922626303325e-06,
+      "loss": 0.0474,
+      "step": 460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6257620584279454e-06,
+      "loss": 0.0881,
+      "step": 461
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5933294470980443e-06,
+      "loss": 0.0475,
+      "step": 462
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5611955811889645e-06,
+      "loss": 0.0473,
+      "step": 463
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5293616026326053e-06,
+      "loss": 0.0143,
+      "step": 464
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4978286427038602e-06,
+      "loss": 0.1228,
+      "step": 465
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4665978219804056e-06,
+      "loss": 0.2635,
+      "step": 466
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.435670250302892e-06,
+      "loss": 0.0668,
+      "step": 467
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.405047026735491e-06,
+      "loss": 0.082,
+      "step": 468
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.3747292395268407e-06,
+      "loss": 0.085,
+      "step": 469
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.344717966071385e-06,
+      "loss": 0.1178,
+      "step": 470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.3150142728710669e-06,
+      "loss": 0.0633,
+      "step": 471
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2856192154974488e-06,
+      "loss": 0.0229,
+      "step": 472
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2565338385541792e-06,
+      "loss": 0.0356,
+      "step": 473
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2277591756398933e-06,
+      "loss": 0.1599,
+      "step": 474
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1992962493114645e-06,
+      "loss": 0.0168,
+      "step": 475
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.171146071047683e-06,
+      "loss": 0.0626,
+      "step": 476
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1433096412132838e-06,
+      "loss": 0.1343,
+      "step": 477
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1157879490234346e-06,
+      "loss": 0.0529,
+      "step": 478
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.088581972508549e-06,
+      "loss": 0.0556,
+      "step": 479
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0616926784795511e-06,
+      "loss": 0.0903,
+      "step": 480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.035121022493506e-06,
+      "loss": 0.0993,
+      "step": 481
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0088679488196695e-06,
+      "loss": 0.0673,
+      "step": 482
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.829343904059342e-07,
+      "loss": 0.018,
+      "step": 483
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.573212688456635e-07,
+      "loss": 0.1005,
+      "step": 484
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.320294943449537e-07,
+      "loss": 0.0859,
+      "step": 485
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.070599656902801e-07,
+      "loss": 0.0361,
+      "step": 486
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.824135702165693e-07,
+      "loss": 0.0256,
+      "step": 487
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.580911837756467e-07,
+      "loss": 0.0652,
+      "step": 488
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.340936707051273e-07,
+      "loss": 0.103,
+      "step": 489
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.10421883797694e-07,
+      "loss": 0.0589,
+      "step": 490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.87076664270795e-07,
+      "loss": 0.1919,
+      "step": 491
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.6405884173675e-07,
+      "loss": 0.1313,
+      "step": 492
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.413692341732582e-07,
+      "loss": 0.0657,
+      "step": 493
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.190086478943459e-07,
+      "loss": 0.1785,
+      "step": 494
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.969778775217007e-07,
+      "loss": 0.1866,
+      "step": 495
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.752777059564431e-07,
+      "loss": 0.0295,
+      "step": 496
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.539089043512914e-07,
+      "loss": 0.0316,
+      "step": 497
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.328722320831737e-07,
+      "loss": 0.0702,
+      "step": 498
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.121684367262271e-07,
+      "loss": 0.0271,
+      "step": 499
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.917982540252442e-07,
+      "loss": 0.0398,
+      "step": 500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.71762407869515e-07,
+      "loss": 0.2131,
+      "step": 501
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.520616102671128e-07,
+      "loss": 0.217,
+      "step": 502
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.326965613195867e-07,
+      "loss": 0.0161,
+      "step": 503
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.136679491970809e-07,
+      "loss": 0.0338,
+      "step": 504
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.949764501138832e-07,
+      "loss": 0.0178,
+      "step": 505
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.766227283043912e-07,
+      "loss": 0.0056,
+      "step": 506
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5860743599951186e-07,
+      "loss": 0.1492,
+      "step": 507
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.4093121340347824e-07,
+      "loss": 0.1143,
+      "step": 508
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.235946886711018e-07,
+      "loss": 0.1318,
+      "step": 509
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.0659847788544926e-07,
+      "loss": 0.0485,
+      "step": 510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.899431850359503e-07,
+      "loss": 0.0155,
+      "step": 511
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.736294019969311e-07,
+      "loss": 0.0895,
+      "step": 512
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5765770850658244e-07,
+      "loss": 0.0913,
+      "step": 513
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.420286721463562e-07,
+      "loss": 0.0394,
+      "step": 514
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2674284832080127e-07,
+      "loss": 0.0125,
+      "step": 515
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.118007802378198e-07,
+      "loss": 0.082,
+      "step": 516
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.972029988893621e-07,
+      "loss": 0.2907,
+      "step": 517
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8295002303256546e-07,
+      "loss": 0.0428,
+      "step": 518
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6904235917131094e-07,
+      "loss": 0.1136,
+      "step": 519
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.554805015382289e-07,
+      "loss": 0.2987,
+      "step": 520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.422649320771331e-07,
+      "loss": 0.1616,
+      "step": 521
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.293961204258932e-07,
+      "loss": 0.1623,
+      "step": 522
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1687452389974829e-07,
+      "loss": 0.0428,
+      "step": 523
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0470058747505516e-07,
+      "loss": 0.0309,
+      "step": 524
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9287474377347238e-07,
+      "loss": 0.0533,
+      "step": 525
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8139741304658566e-07,
+      "loss": 0.0565,
+      "step": 526
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7026900316098217e-07,
+      "loss": 0.0545,
+      "step": 527
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5948990958374543e-07,
+      "loss": 0.1014,
+      "step": 528
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.490605153684066e-07,
+      "loss": 0.2859,
+      "step": 529
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3898119114133192e-07,
+      "loss": 0.2631,
+      "step": 530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.292522950885533e-07,
+      "loss": 0.1341,
+      "step": 531
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1987417294303748e-07,
+      "loss": 0.1001,
+      "step": 532
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1084715797239798e-07,
+      "loss": 0.2334,
+      "step": 533
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0217157096705676e-07,
+      "loss": 0.0793,
+      "step": 534
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.384772022884015e-08,
+      "loss": 0.0426,
+      "step": 535
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.587590156002635e-08,
+      "loss": 0.0174,
+      "step": 536
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.825639825282949e-08,
+      "loss": 0.1224,
+      "step": 537
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.098948107933656e-08,
+      "loss": 0.0296,
+      "step": 538
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.407540828188175e-08,
+      "loss": 0.0831,
+      "step": 539
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7514425563870436e-08,
+      "loss": 0.0264,
+      "step": 540
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1306766081048456e-08,
+      "loss": 0.0982,
+      "step": 541
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.545265043321645e-08,
+      "loss": 0.0393,
+      "step": 542
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.9952286656389506e-08,
+      "loss": 0.0695,
+      "step": 543
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.480587021540527e-08,
+      "loss": 0.1458,
+      "step": 544
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.001358399697618e-08,
+      "loss": 0.0065,
+      "step": 545
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.557559830319245e-08,
+      "loss": 0.1492,
+      "step": 546
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1492070845468005e-08,
+      "loss": 0.094,
+      "step": 547
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7763146738938307e-08,
+      "loss": 0.2497,
+      "step": 548
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4388958497300043e-08,
+      "loss": 0.0833,
+      "step": 549
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1369626028104874e-08,
+      "loss": 0.0528,
+      "step": 550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.705256628499525e-09,
+      "loss": 0.0411,
+      "step": 551
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.39594498140883e-09,
+      "loss": 0.0253,
+      "step": 552
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.4417731521717576e-09,
+      "loss": 0.2015,
+      "step": 553
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.842810585627076e-09,
+      "loss": 0.0077,
+      "step": 554
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5991141036475478e-09,
+      "loss": 0.1521,
+      "step": 555
+    },
+    {
+      "epoch": 1.0,
+      "step": 555,
+      "total_flos": 1196575985664.0,
+      "train_loss": 0.11872713004180172,
+      "train_runtime": 4113.0695,
+      "train_samples_per_second": 1.347,
+      "train_steps_per_second": 0.135
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 555,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "total_flos": 1196575985664.0,
+  "train_batch_size": 10,
+  "trial_name": null,
+  "trial_params": null
+}

CheckGuard Models/wholeimage/bank_no/finetune_lora_llava_mistral.sh

@@ -0,0 +1,43 @@
+#!/bin/bash
+# Use first parameter as GPU IDs, default to "0,1,2,3" if not provided
+GPU_IDS=${1:-0,1,2,3}
+
+
+CUDA_VISIBLE_DEVICES=0,1,2,3 deepspeed --include localhost:"$GPU_IDS" --master_port 29602\
+    llava/train/train_mem.py \
+    --lora_enable True --lora_r 128 --lora_alpha 256 --mm_projector_lr 2e-5 \
+    --deepspeed ./scripts/zero3.json \
+    --model_name_or_path liuhaotian/llava-v1.6-mistral-7b \
+    --version mistral_instruct \
+    --data_path /home/larry5/project/LLaVA-1.6-ft/data/peft/bank_no/bank_no_dataset.json \
+    --image_folder /home/larry5/project/LLaVA-1.6-ft/data/data/ \
+    --vision_tower openai/clip-vit-large-patch14-336 \
+    --mm_projector_type mlp2x_gelu \
+    --mm_vision_select_layer -2 \
+    --mm_use_im_start_end False \
+    --mm_use_im_patch_token False \
+    --mm_patch_merge_type spatial_unpad \
+    --image_aspect_ratio anyres \
+    --group_by_modality_length False \
+    --bf16 False \
+    --fp16 True \
+    --output_dir /home/larry5/project/LLaVA-1.6-ft/scripts_peft/mistral/lora/llava-lora-mistral-r128a256/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model \
+    --num_train_epochs 1 \
+    --per_device_train_batch_size 10 \
+    --per_device_eval_batch_size 1 \
+    --gradient_accumulation_steps 1 \
+    --evaluation_strategy "no" \
+    --save_strategy "steps" \
+    --save_steps 500 \
+    --save_total_limit 5 \
+    --learning_rate 2e-5 \
+    --weight_decay 0. \
+    --warmup_ratio 0.05 \
+    --lr_scheduler_type "cosine" \
+    --logging_steps 1 \
+    --tf32 True \
+    --model_max_length 4096 \
+    --gradient_checkpointing True \
+    --dataloader_num_workers 4 \
+    --lazy_preprocess True \
+    --report_to wandb \

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/README.md

@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: liuhaotian/llava-v1.6-mistral-7b
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.10.0

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "liuhaotian/llava-v1.6-mistral-7b",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "q_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6eea7f9e47bb7d2f074c81e49ccf9648c1394c7fbb7e851b9ac64e47efa2c03b
+size 708924928

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/README.md

@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: liuhaotian/llava-v1.6-mistral-7b
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.10.0

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "liuhaotian/llava-v1.6-mistral-7b",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "q_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:645c293e131efb974f8c218e7d69c93bf50c753554c806d1cf561baa77311585
+size 1417762896

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/global_step500/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3183212032ce3f53bf011c0ea2d72e73d90e4ae83d758f3cb2661945c405d2e
+size 632242

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/global_step500/zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67835d032628ab68661627ea5db2a21c8defdf7306ff43ec6d2d034f2a3add64
+size 4504787266

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/latest

@@ -0,0 +1 @@
+global_step500

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1691c008dc15394c290eec92c6d96f1d3cc3096220a9fdad0f2210c4f3699fd5
+size 14244

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b37e2b05185c6152f2a40fb75a789b697d3a87176492c5cbb481ba82522c2163
+size 1064

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/special_tokens_map.json

@@ -0,0 +1,24 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<unk>",
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055
+size 493443

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/tokenizer_config.json

@@ -0,0 +1,44 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [],
+  "bos_token": "<s>",
+  "chat_template": "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token}}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": true,
+  "model_max_length": 4096,
+  "pad_token": "<unk>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/trainer_state.json

@@ -0,0 +1,3021 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9242144177449169,
+  "eval_steps": 500,
+  "global_step": 500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.142857142857143e-07,
+      "loss": 0.6789,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4285714285714286e-06,
+      "loss": 0.8481,
+      "step": 2
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1428571428571427e-06,
+      "loss": 0.663,
+      "step": 3
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8571428571428573e-06,
+      "loss": 0.679,
+      "step": 4
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5714285714285718e-06,
+      "loss": 1.0166,
+      "step": 5
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2857142857142855e-06,
+      "loss": 0.4693,
+      "step": 6
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5e-06,
+      "loss": 0.4891,
+      "step": 7
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.7142857142857145e-06,
+      "loss": 0.5523,
+      "step": 8
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.4285714285714295e-06,
+      "loss": 0.2909,
+      "step": 9
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.1428571428571436e-06,
+      "loss": 0.2598,
+      "step": 10
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.857142857142858e-06,
+      "loss": 0.2532,
+      "step": 11
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.857142857142858e-06,
+      "loss": 0.4867,
+      "step": 12
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 0.4145,
+      "step": 13
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 0.3161,
+      "step": 14
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.285714285714288e-06,
+      "loss": 0.1836,
+      "step": 15
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-05,
+      "loss": 0.3355,
+      "step": 16
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0714285714285714e-05,
+      "loss": 0.2286,
+      "step": 17
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 0.3594,
+      "step": 18
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2142857142857142e-05,
+      "loss": 0.2981,
+      "step": 19
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2857142857142859e-05,
+      "loss": 0.3021,
+      "step": 20
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3571428571428574e-05,
+      "loss": 0.3866,
+      "step": 21
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4285714285714287e-05,
+      "loss": 0.2409,
+      "step": 22
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.1397,
+      "step": 23
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.5714285714285715e-05,
+      "loss": 0.1416,
+      "step": 24
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.642857142857143e-05,
+      "loss": 0.1838,
+      "step": 25
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7142857142857142e-05,
+      "loss": 0.1505,
+      "step": 26
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.785714285714286e-05,
+      "loss": 0.3278,
+      "step": 27
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8571428571428575e-05,
+      "loss": 0.2567,
+      "step": 28
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.928571428571429e-05,
+      "loss": 0.1218,
+      "step": 29
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2e-05,
+      "loss": 0.2288,
+      "step": 30
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9999812486015525e-05,
+      "loss": 0.1348,
+      "step": 31
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9999249951094388e-05,
+      "loss": 0.3734,
+      "step": 32
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999831241633323e-05,
+      "loss": 0.3169,
+      "step": 33
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9996999916892222e-05,
+      "loss": 0.1066,
+      "step": 34
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9995312501993765e-05,
+      "loss": 0.4434,
+      "step": 35
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9993250234920638e-05,
+      "loss": 0.198,
+      "step": 36
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9990813193013625e-05,
+      "loss": 0.115,
+      "step": 37
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9988001467668613e-05,
+      "loss": 0.2676,
+      "step": 38
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9984815164333163e-05,
+      "loss": 0.2201,
+      "step": 39
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9981254402502568e-05,
+      "loss": 0.1945,
+      "step": 40
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.997731931571535e-05,
+      "loss": 0.1391,
+      "step": 41
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9973010051548274e-05,
+      "loss": 0.2697,
+      "step": 42
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968326771610797e-05,
+      "loss": 0.1562,
+      "step": 43
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9963269651539018e-05,
+      "loss": 0.2204,
+      "step": 44
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9957838880989076e-05,
+      "loss": 0.2729,
+      "step": 45
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952034663630064e-05,
+      "loss": 0.441,
+      "step": 46
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952034663630064e-05,
+      "loss": 0.1401,
+      "step": 47
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9945857217136365e-05,
+      "loss": 0.3727,
+      "step": 48
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9939306773179498e-05,
+      "loss": 0.3269,
+      "step": 49
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9932383577419432e-05,
+      "loss": 0.0801,
+      "step": 50
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9925087889495374e-05,
+      "loss": 0.2772,
+      "step": 51
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9917419983016025e-05,
+      "loss": 0.2253,
+      "step": 52
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9909380145549325e-05,
+      "loss": 0.2318,
+      "step": 53
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9900968678611664e-05,
+      "loss": 0.1809,
+      "step": 54
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.989218589765658e-05,
+      "loss": 0.1155,
+      "step": 55
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9883032132062926e-05,
+      "loss": 0.2356,
+      "step": 56
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9873507725122505e-05,
+      "loss": 0.1194,
+      "step": 57
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9863613034027224e-05,
+      "loss": 0.3272,
+      "step": 58
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.985334842985567e-05,
+      "loss": 0.183,
+      "step": 59
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9842714297559212e-05,
+      "loss": 0.1217,
+      "step": 60
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9831711035947552e-05,
+      "loss": 0.1388,
+      "step": 61
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9820339057673773e-05,
+      "loss": 0.2112,
+      "step": 62
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9808598789218866e-05,
+      "loss": 0.0917,
+      "step": 63
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.979649067087574e-05,
+      "loss": 0.1585,
+      "step": 64
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9784015156732693e-05,
+      "loss": 0.1446,
+      "step": 65
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.97711727146564e-05,
+      "loss": 0.3511,
+      "step": 66
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9757963826274357e-05,
+      "loss": 0.1019,
+      "step": 67
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9744388986956824e-05,
+      "loss": 0.1165,
+      "step": 68
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.973044870579824e-05,
+      "loss": 0.2189,
+      "step": 69
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.971614350559814e-05,
+      "loss": 0.1254,
+      "step": 70
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.970147392284154e-05,
+      "loss": 0.0627,
+      "step": 71
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9686440507678827e-05,
+      "loss": 0.0952,
+      "step": 72
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.967104382390511e-05,
+      "loss": 0.1867,
+      "step": 73
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9655284448939094e-05,
+      "loss": 0.2003,
+      "step": 74
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9639162973801426e-05,
+      "loss": 0.1188,
+      "step": 75
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9622680003092503e-05,
+      "loss": 0.1111,
+      "step": 76
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.960583615496984e-05,
+      "loss": 0.1203,
+      "step": 77
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9588632061124837e-05,
+      "loss": 0.1599,
+      "step": 78
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9571068366759143e-05,
+      "loss": 0.209,
+      "step": 79
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9553145730560415e-05,
+      "loss": 0.2183,
+      "step": 80
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.953486482467764e-05,
+      "loss": 0.1351,
+      "step": 81
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.951622633469592e-05,
+      "loss": 0.128,
+      "step": 82
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9497230959610757e-05,
+      "loss": 0.2241,
+      "step": 83
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9477879411801843e-05,
+      "loss": 0.0991,
+      "step": 84
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9458172417006347e-05,
+      "loss": 0.1165,
+      "step": 85
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9438110714291697e-05,
+      "loss": 0.0792,
+      "step": 86
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9417695056027847e-05,
+      "loss": 0.121,
+      "step": 87
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9396926207859085e-05,
+      "loss": 0.2727,
+      "step": 88
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9375804948675308e-05,
+      "loss": 0.1947,
+      "step": 89
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.935433207058281e-05,
+      "loss": 0.2155,
+      "step": 90
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.933250837887457e-05,
+      "loss": 0.0525,
+      "step": 91
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9310334692000077e-05,
+      "loss": 0.2401,
+      "step": 92
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9287811841534598e-05,
+      "loss": 0.0743,
+      "step": 93
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9264940672148018e-05,
+      "loss": 0.1659,
+      "step": 94
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9241722041573166e-05,
+      "loss": 0.1184,
+      "step": 95
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9218156820573618e-05,
+      "loss": 0.1207,
+      "step": 96
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9194245892911077e-05,
+      "loss": 0.1292,
+      "step": 97
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.916999015531221e-05,
+      "loss": 0.2059,
+      "step": 98
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9145390517435013e-05,
+      "loss": 0.1682,
+      "step": 99
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9120447901834708e-05,
+      "loss": 0.1403,
+      "step": 100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9095163243929143e-05,
+      "loss": 0.1752,
+      "step": 101
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.906953749196371e-05,
+      "loss": 0.1616,
+      "step": 102
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9043571606975776e-05,
+      "loss": 0.1127,
+      "step": 103
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.901726656275866e-05,
+      "loss": 0.2236,
+      "step": 104
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8990623345825084e-05,
+      "loss": 0.2308,
+      "step": 105
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8963642955370203e-05,
+      "loss": 0.1739,
+      "step": 106
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8936326403234125e-05,
+      "loss": 0.1762,
+      "step": 107
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.890867471386395e-05,
+      "loss": 0.1457,
+      "step": 108
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.888068892427538e-05,
+      "loss": 0.2768,
+      "step": 109
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8852370084013783e-05,
+      "loss": 0.1389,
+      "step": 110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.882371925511488e-05,
+      "loss": 0.2747,
+      "step": 111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.879473751206489e-05,
+      "loss": 0.0542,
+      "step": 112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8765425941760237e-05,
+      "loss": 0.1414,
+      "step": 113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8735785643466786e-05,
+      "loss": 0.2482,
+      "step": 114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8705817728778626e-05,
+      "loss": 0.1602,
+      "step": 115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.867552332157637e-05,
+      "loss": 0.1342,
+      "step": 116
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8644903557985027e-05,
+      "loss": 0.077,
+      "step": 117
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8613959586331364e-05,
+      "loss": 0.0818,
+      "step": 118
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8582692567100866e-05,
+      "loss": 0.1443,
+      "step": 119
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855110367289421e-05,
+      "loss": 0.1148,
+      "step": 120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851919408838327e-05,
+      "loss": 0.1661,
+      "step": 121
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8486965010266726e-05,
+      "loss": 0.1676,
+      "step": 122
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845441764722514e-05,
+      "loss": 0.1288,
+      "step": 123
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.842155321987566e-05,
+      "loss": 0.0725,
+      "step": 124
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8388372960726228e-05,
+      "loss": 0.1258,
+      "step": 125
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8354878114129368e-05,
+      "loss": 0.068,
+      "step": 126
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8321069936235503e-05,
+      "loss": 0.1698,
+      "step": 127
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8286949694945864e-05,
+      "loss": 0.2038,
+      "step": 128
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8252518669864935e-05,
+      "loss": 0.0274,
+      "step": 129
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.821777815225245e-05,
+      "loss": 0.0564,
+      "step": 130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8182729444974993e-05,
+      "loss": 0.1182,
+      "step": 131
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8147373862457107e-05,
+      "loss": 0.3175,
+      "step": 132
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8111712730632024e-05,
+      "loss": 0.1017,
+      "step": 133
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.807574738689193e-05,
+      "loss": 0.3348,
+      "step": 134
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8039479180037803e-05,
+      "loss": 0.3129,
+      "step": 135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.800290947022884e-05,
+      "loss": 0.1095,
+      "step": 136
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7966039628931447e-05,
+      "loss": 0.1922,
+      "step": 137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7928871038867785e-05,
+      "loss": 0.1022,
+      "step": 138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.789140509396394e-05,
+      "loss": 0.2318,
+      "step": 139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7853643199297632e-05,
+      "loss": 0.2374,
+      "step": 140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7815586771045535e-05,
+      "loss": 0.1194,
+      "step": 141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.777723723643014e-05,
+      "loss": 0.1914,
+      "step": 142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.773859603366626e-05,
+      "loss": 0.0431,
+      "step": 143
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.769966461190707e-05,
+      "loss": 0.081,
+      "step": 144
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.766044443118978e-05,
+      "loss": 0.2162,
+      "step": 145
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.762093696238086e-05,
+      "loss": 0.1151,
+      "step": 146
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7581143687120877e-05,
+      "loss": 0.184,
+      "step": 147
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7541066097768965e-05,
+      "loss": 0.1963,
+      "step": 148
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.750070569734681e-05,
+      "loss": 0.1318,
+      "step": 149
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7460063999482314e-05,
+      "loss": 0.1163,
+      "step": 150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7419142528352815e-05,
+      "loss": 0.1013,
+      "step": 151
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.737794281862794e-05,
+      "loss": 0.0957,
+      "step": 152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7336466415412028e-05,
+      "loss": 0.2023,
+      "step": 153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.729471487418621e-05,
+      "loss": 0.1398,
+      "step": 154
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7252689760750053e-05,
+      "loss": 0.1238,
+      "step": 155
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.721039265116285e-05,
+      "loss": 0.2201,
+      "step": 156
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7167825131684516e-05,
+      "loss": 0.0698,
+      "step": 157
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7124988798716084e-05,
+      "loss": 0.0312,
+      "step": 158
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7081885258739846e-05,
+      "loss": 0.1443,
+      "step": 159
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7038516128259118e-05,
+      "loss": 0.1349,
+      "step": 160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6994883033737582e-05,
+      "loss": 0.0751,
+      "step": 161
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.695098761153832e-05,
+      "loss": 0.0543,
+      "step": 162
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6906831507862446e-05,
+      "loss": 0.0533,
+      "step": 163
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.686241637868734e-05,
+      "loss": 0.1328,
+      "step": 164
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6817743889704564e-05,
+      "loss": 0.3057,
+      "step": 165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6772815716257414e-05,
+      "loss": 0.1642,
+      "step": 166
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.672763354327804e-05,
+      "loss": 0.1479,
+      "step": 167
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6682199065224307e-05,
+      "loss": 0.1163,
+      "step": 168
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6636513986016215e-05,
+      "loss": 0.0395,
+      "step": 169
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6590580018972012e-05,
+      "loss": 0.0456,
+      "step": 170
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6544398886743934e-05,
+      "loss": 0.2018,
+      "step": 171
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.64979723212536e-05,
+      "loss": 0.1655,
+      "step": 172
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6451302063627067e-05,
+      "loss": 0.1805,
+      "step": 173
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6404389864129533e-05,
+      "loss": 0.2445,
+      "step": 174
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6357237482099682e-05,
+      "loss": 0.134,
+      "step": 175
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6309846685883726e-05,
+      "loss": 0.0976,
+      "step": 176
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6262219252769065e-05,
+      "loss": 0.0984,
+      "step": 177
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.621435696891765e-05,
+      "loss": 0.0495,
+      "step": 178
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6166261629298996e-05,
+      "loss": 0.1005,
+      "step": 179
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6117935037622848e-05,
+      "loss": 0.1399,
+      "step": 180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.606937900627157e-05,
+      "loss": 0.2105,
+      "step": 181
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.6020595356232137e-05,
+      "loss": 0.142,
+      "step": 182
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5971585917027864e-05,
+      "loss": 0.0791,
+      "step": 183
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5922352526649803e-05,
+      "loss": 0.2,
+      "step": 184
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.587289703148779e-05,
+      "loss": 0.1317,
+      "step": 185
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5823221286261217e-05,
+      "loss": 0.1656,
+      "step": 186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5773327153949465e-05,
+      "loss": 0.3358,
+      "step": 187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.572321650572205e-05,
+      "loss": 0.2216,
+      "step": 188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.567289122086843e-05,
+      "loss": 0.0937,
+      "step": 189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5622353186727542e-05,
+      "loss": 0.0995,
+      "step": 190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.557160429861702e-05,
+      "loss": 0.2324,
+      "step": 191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5520646459762102e-05,
+      "loss": 0.2847,
+      "step": 192
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5469481581224274e-05,
+      "loss": 0.1242,
+      "step": 193
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5418111581829575e-05,
+      "loss": 0.1771,
+      "step": 194
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.536653838809667e-05,
+      "loss": 0.2115,
+      "step": 195
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.531476393416456e-05,
+      "loss": 0.074,
+      "step": 196
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5262790161720082e-05,
+      "loss": 0.0893,
+      "step": 197
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5210619019925066e-05,
+      "loss": 0.0644,
+      "step": 198
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5158252465343242e-05,
+      "loss": 0.2146,
+      "step": 199
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5105692461866874e-05,
+      "loss": 0.2579,
+      "step": 200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.50529409806431e-05,
+      "loss": 0.0806,
+      "step": 201
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.0806,
+      "step": 202
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4946871505372426e-05,
+      "loss": 0.132,
+      "step": 203
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4893557489227518e-05,
+      "loss": 0.1438,
+      "step": 204
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4840059950989992e-05,
+      "loss": 0.1703,
+      "step": 205
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.478638089696716e-05,
+      "loss": 0.0903,
+      "step": 206
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4732522340273686e-05,
+      "loss": 0.1515,
+      "step": 207
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.467848630075608e-05,
+      "loss": 0.2156,
+      "step": 208
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4624274804916958e-05,
+      "loss": 0.0783,
+      "step": 209
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.456988988583904e-05,
+      "loss": 0.1432,
+      "step": 210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4515333583108896e-05,
+      "loss": 0.1716,
+      "step": 211
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4460607942740468e-05,
+      "loss": 0.2328,
+      "step": 212
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4405715017098333e-05,
+      "loss": 0.1317,
+      "step": 213
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4350656864820733e-05,
+      "loss": 0.097,
+      "step": 214
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4295435550742372e-05,
+      "loss": 0.1547,
+      "step": 215
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4240053145816968e-05,
+      "loss": 0.0737,
+      "step": 216
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4184511727039612e-05,
+      "loss": 0.0926,
+      "step": 217
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4128813377368851e-05,
+      "loss": 0.0824,
+      "step": 218
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4072960185648576e-05,
+      "loss": 0.1236,
+      "step": 219
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.4016954246529697e-05,
+      "loss": 0.157,
+      "step": 220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.396079766039157e-05,
+      "loss": 0.1241,
+      "step": 221
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3904492533263243e-05,
+      "loss": 0.1243,
+      "step": 222
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3848040976744459e-05,
+      "loss": 0.1429,
+      "step": 223
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3791445107926478e-05,
+      "loss": 0.0321,
+      "step": 224
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3734707049312674e-05,
+      "loss": 0.0398,
+      "step": 225
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3677828928738934e-05,
+      "loss": 0.2625,
+      "step": 226
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3620812879293864e-05,
+      "loss": 0.0926,
+      "step": 227
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3563661039238785e-05,
+      "loss": 0.06,
+      "step": 228
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3506375551927546e-05,
+      "loss": 0.2397,
+      "step": 229
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3448958565726144e-05,
+      "loss": 0.157,
+      "step": 230
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3391412233932148e-05,
+      "loss": 0.1105,
+      "step": 231
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3333738714693958e-05,
+      "loss": 0.0877,
+      "step": 232
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3275940170929845e-05,
+      "loss": 0.1821,
+      "step": 233
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3218018770246858e-05,
+      "loss": 0.0166,
+      "step": 234
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3159976684859528e-05,
+      "loss": 0.118,
+      "step": 235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.3101816091508389e-05,
+      "loss": 0.2289,
+      "step": 236
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.3043539171378362e-05,
+      "loss": 0.0518,
+      "step": 237
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2985148110016947e-05,
+      "loss": 0.1012,
+      "step": 238
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.292664509725226e-05,
+      "loss": 0.2009,
+      "step": 239
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2868032327110904e-05,
+      "loss": 0.252,
+      "step": 240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2809311997735697e-05,
+      "loss": 0.2044,
+      "step": 241
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2750486311303218e-05,
+      "loss": 0.1908,
+      "step": 242
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2691557473941246e-05,
+      "loss": 0.3064,
+      "step": 243
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2632527695645993e-05,
+      "loss": 0.091,
+      "step": 244
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.257339919019925e-05,
+      "loss": 0.0606,
+      "step": 245
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2514174175085346e-05,
+      "loss": 0.147,
+      "step": 246
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2454854871407993e-05,
+      "loss": 0.2029,
+      "step": 247
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.239544350380699e-05,
+      "loss": 0.0851,
+      "step": 248
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2335942300374788e-05,
+      "loss": 0.0904,
+      "step": 249
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2276353492572937e-05,
+      "loss": 0.0721,
+      "step": 250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2216679315148388e-05,
+      "loss": 0.1488,
+      "step": 251
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.2156922006049703e-05,
+      "loss": 0.1927,
+      "step": 252
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.2097083806343104e-05,
+      "loss": 0.029,
+      "step": 253
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.2037166960128443e-05,
+      "loss": 0.0301,
+      "step": 254
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1977173714455034e-05,
+      "loss": 0.1231,
+      "step": 255
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1917106319237386e-05,
+      "loss": 0.0348,
+      "step": 256
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1856967027170818e-05,
+      "loss": 0.0869,
+      "step": 257
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1796758093646989e-05,
+      "loss": 0.1164,
+      "step": 258
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1736481776669307e-05,
+      "loss": 0.0388,
+      "step": 259
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1676140336768236e-05,
+      "loss": 0.0433,
+      "step": 260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.161573603691655e-05,
+      "loss": 0.1996,
+      "step": 261
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1555271142444433e-05,
+      "loss": 0.2182,
+      "step": 262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1494747920954545e-05,
+      "loss": 0.0509,
+      "step": 263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1434168642236964e-05,
+      "loss": 0.1078,
+      "step": 264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1373535578184083e-05,
+      "loss": 0.0412,
+      "step": 265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1312851002705383e-05,
+      "loss": 0.2425,
+      "step": 266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1252117191642175e-05,
+      "loss": 0.1119,
+      "step": 267
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1191336422682237e-05,
+      "loss": 0.0455,
+      "step": 268
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1130510975274408e-05,
+      "loss": 0.2613,
+      "step": 269
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1069643130543084e-05,
+      "loss": 0.0651,
+      "step": 270
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1008735171202685e-05,
+      "loss": 0.1155,
+      "step": 271
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0947789381472035e-05,
+      "loss": 0.0661,
+      "step": 272
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0886808046988716e-05,
+      "loss": 0.0881,
+      "step": 273
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0825793454723325e-05,
+      "loss": 0.1123,
+      "step": 274
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0764747892893724e-05,
+      "loss": 0.14,
+      "step": 275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0703673650879219e-05,
+      "loss": 0.0889,
+      "step": 276
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0642573019134703e-05,
+      "loss": 0.1333,
+      "step": 277
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0581448289104759e-05,
+      "loss": 0.0608,
+      "step": 278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0520301753137725e-05,
+      "loss": 0.2882,
+      "step": 279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.045913570439972e-05,
+      "loss": 0.0661,
+      "step": 280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0397952436788643e-05,
+      "loss": 0.107,
+      "step": 281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0336754244848156e-05,
+      "loss": 0.0499,
+      "step": 282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0275543423681622e-05,
+      "loss": 0.237,
+      "step": 283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0214322268866033e-05,
+      "loss": 0.0301,
+      "step": 284
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0153093076365923e-05,
+      "loss": 0.0904,
+      "step": 285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0091858142447266e-05,
+      "loss": 0.0165,
+      "step": 286
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0030619763591348e-05,
+      "loss": 0.0791,
+      "step": 287
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.969380236408656e-06,
+      "loss": 0.1997,
+      "step": 288
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.908141857552737e-06,
+      "loss": 0.0155,
+      "step": 289
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.846906923634079e-06,
+      "loss": 0.0457,
+      "step": 290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.785677731133972e-06,
+      "loss": 0.0203,
+      "step": 291
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.724456576318383e-06,
+      "loss": 0.2384,
+      "step": 292
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.663245755151847e-06,
+      "loss": 0.1459,
+      "step": 293
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.602047563211359e-06,
+      "loss": 0.2249,
+      "step": 294
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.540864295600282e-06,
+      "loss": 0.037,
+      "step": 295
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.479698246862277e-06,
+      "loss": 0.145,
+      "step": 296
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.418551710895243e-06,
+      "loss": 0.1501,
+      "step": 297
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.3574269808653e-06,
+      "loss": 0.0727,
+      "step": 298
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.296326349120786e-06,
+      "loss": 0.0992,
+      "step": 299
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.23525210710628e-06,
+      "loss": 0.2516,
+      "step": 300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.174206545276678e-06,
+      "loss": 0.0628,
+      "step": 301
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.113191953011287e-06,
+      "loss": 0.132,
+      "step": 302
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.052210618527966e-06,
+      "loss": 0.0908,
+      "step": 303
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.991264828797319e-06,
+      "loss": 0.1432,
+      "step": 304
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.93035686945692e-06,
+      "loss": 0.0493,
+      "step": 305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.869489024725595e-06,
+      "loss": 0.0578,
+      "step": 306
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.808663577317765e-06,
+      "loss": 0.0909,
+      "step": 307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.747882808357828e-06,
+      "loss": 0.0646,
+      "step": 308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.687148997294622e-06,
+      "loss": 0.1308,
+      "step": 309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.626464421815919e-06,
+      "loss": 0.0729,
+      "step": 310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.565831357763039e-06,
+      "loss": 0.1871,
+      "step": 311
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.505252079045459e-06,
+      "loss": 0.1577,
+      "step": 312
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.444728857555572e-06,
+      "loss": 0.1844,
+      "step": 313
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.384263963083453e-06,
+      "loss": 0.1673,
+      "step": 314
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.323859663231768e-06,
+      "loss": 0.1898,
+      "step": 315
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.263518223330698e-06,
+      "loss": 0.1106,
+      "step": 316
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.203241906353014e-06,
+      "loss": 0.0476,
+      "step": 317
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.143032972829184e-06,
+      "loss": 0.1432,
+      "step": 318
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.082893680762619e-06,
+      "loss": 0.0249,
+      "step": 319
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.022826285544967e-06,
+      "loss": 0.0762,
+      "step": 320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.962833039871562e-06,
+      "loss": 0.1468,
+      "step": 321
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.902916193656898e-06,
+      "loss": 0.0272,
+      "step": 322
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.843077993950302e-06,
+      "loss": 0.0495,
+      "step": 323
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.783320684851613e-06,
+      "loss": 0.1958,
+      "step": 324
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.72364650742707e-06,
+      "loss": 0.0869,
+      "step": 325
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.664057699625215e-06,
+      "loss": 0.2957,
+      "step": 326
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.604556496193015e-06,
+      "loss": 0.0833,
+      "step": 327
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.545145128592009e-06,
+      "loss": 0.0978,
+      "step": 328
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.485825824914658e-06,
+      "loss": 0.1941,
+      "step": 329
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.426600809800753e-06,
+      "loss": 0.0384,
+      "step": 330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.367472304354011e-06,
+      "loss": 0.0872,
+      "step": 331
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.308442526058757e-06,
+      "loss": 0.1051,
+      "step": 332
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.249513688696786e-06,
+      "loss": 0.0918,
+      "step": 333
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.190688002264308e-06,
+      "loss": 0.2169,
+      "step": 334
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.131967672889101e-06,
+      "loss": 0.1647,
+      "step": 335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.073354902747742e-06,
+      "loss": 0.0585,
+      "step": 336
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.014851889983058e-06,
+      "loss": 0.1743,
+      "step": 337
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.956460828621641e-06,
+      "loss": 0.3001,
+      "step": 338
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.898183908491617e-06,
+      "loss": 0.0977,
+      "step": 339
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.840023315140476e-06,
+      "loss": 0.0549,
+      "step": 340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.781981229753145e-06,
+      "loss": 0.0738,
+      "step": 341
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.7240598290701585e-06,
+      "loss": 0.027,
+      "step": 342
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.666261285306048e-06,
+      "loss": 0.0647,
+      "step": 343
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.608587766067853e-06,
+      "loss": 0.0531,
+      "step": 344
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.551041434273862e-06,
+      "loss": 0.0582,
+      "step": 345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.4936244480724575e-06,
+      "loss": 0.2357,
+      "step": 346
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.4363389607612204e-06,
+      "loss": 0.0614,
+      "step": 347
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.379187120706138e-06,
+      "loss": 0.1516,
+      "step": 348
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.322171071261071e-06,
+      "loss": 0.2906,
+      "step": 349
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.265292950687329e-06,
+      "loss": 0.0402,
+      "step": 350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.208554892073528e-06,
+      "loss": 0.0895,
+      "step": 351
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.151959023255545e-06,
+      "loss": 0.109,
+      "step": 352
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.095507466736763e-06,
+      "loss": 0.1338,
+      "step": 353
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.039202339608432e-06,
+      "loss": 0.0541,
+      "step": 354
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.983045753470308e-06,
+      "loss": 0.0614,
+      "step": 355
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.927039814351426e-06,
+      "loss": 0.2844,
+      "step": 356
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.871186622631155e-06,
+      "loss": 0.1412,
+      "step": 357
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.815488272960388e-06,
+      "loss": 0.0575,
+      "step": 358
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.759946854183036e-06,
+      "loss": 0.1047,
+      "step": 359
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.704564449257635e-06,
+      "loss": 0.2065,
+      "step": 360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.649343135179271e-06,
+      "loss": 0.0995,
+      "step": 361
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.59428498290167e-06,
+      "loss": 0.1517,
+      "step": 362
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.539392057259536e-06,
+      "loss": 0.1122,
+      "step": 363
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.484666416891109e-06,
+      "loss": 0.0992,
+      "step": 364
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.430110114160965e-06,
+      "loss": 0.1303,
+      "step": 365
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.375725195083046e-06,
+      "loss": 0.1192,
+      "step": 366
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.321513699243924e-06,
+      "loss": 0.0991,
+      "step": 367
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.267477659726319e-06,
+      "loss": 0.077,
+      "step": 368
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.213619103032845e-06,
+      "loss": 0.1052,
+      "step": 369
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.159940049010015e-06,
+      "loss": 0.2359,
+      "step": 370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 5.106442510772489e-06,
+      "loss": 0.0501,
+      "step": 371
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 5.053128494627578e-06,
+      "loss": 0.0803,
+      "step": 372
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 5.000000000000003e-06,
+      "loss": 0.2073,
+      "step": 373
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.947059019356904e-06,
+      "loss": 0.0479,
+      "step": 374
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.89430753813313e-06,
+      "loss": 0.125,
+      "step": 375
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.8417475346567635e-06,
+      "loss": 0.0715,
+      "step": 376
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.78938098007494e-06,
+      "loss": 0.0242,
+      "step": 377
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.737209838279923e-06,
+      "loss": 0.1242,
+      "step": 378
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.685236065835443e-06,
+      "loss": 0.1771,
+      "step": 379
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.633461611903336e-06,
+      "loss": 0.1037,
+      "step": 380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.581888418170429e-06,
+      "loss": 0.0733,
+      "step": 381
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.530518418775734e-06,
+      "loss": 0.0565,
+      "step": 382
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.479353540237903e-06,
+      "loss": 0.1092,
+      "step": 383
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.4283957013829845e-06,
+      "loss": 0.0371,
+      "step": 384
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3776468132724605e-06,
+      "loss": 0.1105,
+      "step": 385
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.327108779131573e-06,
+      "loss": 0.1856,
+      "step": 386
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.276783494277954e-06,
+      "loss": 0.1237,
+      "step": 387
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.226672846050538e-06,
+      "loss": 0.2521,
+      "step": 388
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.176778713738787e-06,
+      "loss": 0.0565,
+      "step": 389
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.127102968512214e-06,
+      "loss": 0.0518,
+      "step": 390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.077647473350201e-06,
+      "loss": 0.0735,
+      "step": 391
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.028414082972141e-06,
+      "loss": 0.0786,
+      "step": 392
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.9794046437678705e-06,
+      "loss": 0.025,
+      "step": 393
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.930620993728434e-06,
+      "loss": 0.2235,
+      "step": 394
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.882064962377154e-06,
+      "loss": 0.1307,
+      "step": 395
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.83373837070101e-06,
+      "loss": 0.0224,
+      "step": 396
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7856430310823546e-06,
+      "loss": 0.1109,
+      "step": 397
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.737780747230941e-06,
+      "loss": 0.0624,
+      "step": 398
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.6901533141162804e-06,
+      "loss": 0.055,
+      "step": 399
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.6427625179003223e-06,
+      "loss": 0.2079,
+      "step": 400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.595610135870472e-06,
+      "loss": 0.2215,
+      "step": 401
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.548697936372937e-06,
+      "loss": 0.1016,
+      "step": 402
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5020276787464058e-06,
+      "loss": 0.1229,
+      "step": 403
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.455601113256073e-06,
+      "loss": 0.0759,
+      "step": 404
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.4094199810279926e-06,
+      "loss": 0.1667,
+      "step": 405
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.3634860139837877e-06,
+      "loss": 0.048,
+      "step": 406
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.317800934775696e-06,
+      "loss": 0.0543,
+      "step": 407
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2723664567219627e-06,
+      "loss": 0.1656,
+      "step": 408
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.2271842837425917e-06,
+      "loss": 0.0409,
+      "step": 409
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.1822561102954373e-06,
+      "loss": 0.1173,
+      "step": 410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.1375836213126653e-06,
+      "loss": 0.0964,
+      "step": 411
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0931684921375572e-06,
+      "loss": 0.0432,
+      "step": 412
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0490123884616795e-06,
+      "loss": 0.1451,
+      "step": 413
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.0051169662624224e-06,
+      "loss": 0.1226,
+      "step": 414
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.9614838717408866e-06,
+      "loss": 0.096,
+      "step": 415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.918114741260156e-06,
+      "loss": 0.1152,
+      "step": 416
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8750112012839215e-06,
+      "loss": 0.0575,
+      "step": 417
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8321748683154893e-06,
+      "loss": 0.097,
+      "step": 418
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7896073488371535e-06,
+      "loss": 0.0513,
+      "step": 419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7473102392499517e-06,
+      "loss": 0.0566,
+      "step": 420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7052851258137936e-06,
+      "loss": 0.0193,
+      "step": 421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.663533584587974e-06,
+      "loss": 0.1507,
+      "step": 422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.622057181372063e-06,
+      "loss": 0.0208,
+      "step": 423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.580857471647186e-06,
+      "loss": 0.0893,
+      "step": 424
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.539936000517689e-06,
+      "loss": 0.0988,
+      "step": 425
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4992943026531935e-06,
+      "loss": 0.0368,
+      "step": 426
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4589339022310386e-06,
+      "loss": 0.0911,
+      "step": 427
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4188563128791255e-06,
+      "loss": 0.1093,
+      "step": 428
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.379063037619146e-06,
+      "loss": 0.0717,
+      "step": 429
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.339555568810221e-06,
+      "loss": 0.1486,
+      "step": 430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.300335388092929e-06,
+      "loss": 0.1174,
+      "step": 431
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.261403966333742e-06,
+      "loss": 0.2022,
+      "step": 432
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.2227627635698624e-06,
+      "loss": 0.0376,
+      "step": 433
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1844132289544684e-06,
+      "loss": 0.3022,
+      "step": 434
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1463568007023706e-06,
+      "loss": 0.0121,
+      "step": 435
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.1085949060360654e-06,
+      "loss": 0.1441,
+      "step": 436
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0711289611322204e-06,
+      "loss": 0.0457,
+      "step": 437
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0339603710685574e-06,
+      "loss": 0.0324,
+      "step": 438
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9970905297711606e-06,
+      "loss": 0.045,
+      "step": 439
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9605208199621993e-06,
+      "loss": 0.0644,
+      "step": 440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.924252613108073e-06,
+      "loss": 0.0743,
+      "step": 441
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8882872693679787e-06,
+      "loss": 0.054,
+      "step": 442
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8526261375428955e-06,
+      "loss": 0.1679,
+      "step": 443
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8172705550250093e-06,
+      "loss": 0.0666,
+      "step": 444
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7822218477475496e-06,
+      "loss": 0.2,
+      "step": 445
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7474813301350668e-06,
+      "loss": 0.1191,
+      "step": 446
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7130503050541368e-06,
+      "loss": 0.1166,
+      "step": 447
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6789300637645e-06,
+      "loss": 0.0089,
+      "step": 448
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6451218858706374e-06,
+      "loss": 0.0848,
+      "step": 449
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6116270392737753e-06,
+      "loss": 0.1263,
+      "step": 450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.578446780124344e-06,
+      "loss": 0.1338,
+      "step": 451
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5455823527748626e-06,
+      "loss": 0.0566,
+      "step": 452
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5130349897332764e-06,
+      "loss": 0.0618,
+      "step": 453
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4808059116167306e-06,
+      "loss": 0.0259,
+      "step": 454
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4488963271057943e-06,
+      "loss": 0.1682,
+      "step": 455
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4173074328991376e-06,
+      "loss": 0.0967,
+      "step": 456
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3860404136686411e-06,
+      "loss": 0.0799,
+      "step": 457
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.355096442014977e-06,
+      "loss": 0.1426,
+      "step": 458
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.3244766784236307e-06,
+      "loss": 0.1401,
+      "step": 459
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.294182271221377e-06,
+      "loss": 0.0526,
+      "step": 460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2642143565332154e-06,
+      "loss": 0.1516,
+      "step": 461
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2345740582397647e-06,
+      "loss": 0.0326,
+      "step": 462
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.2052624879351105e-06,
+      "loss": 0.0517,
+      "step": 463
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.176280744885121e-06,
+      "loss": 0.094,
+      "step": 464
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1476299159862204e-06,
+      "loss": 0.0684,
+      "step": 465
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1193110757246251e-06,
+      "loss": 0.0845,
+      "step": 466
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.09132528613605e-06,
+      "loss": 0.1105,
+      "step": 467
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0636735967658785e-06,
+      "loss": 0.0947,
+      "step": 468
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0363570446297999e-06,
+      "loss": 0.0685,
+      "step": 469
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0093766541749206e-06,
+      "loss": 0.0902,
+      "step": 470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.827334372413444e-07,
+      "loss": 0.0257,
+      "step": 471
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.564283930242258e-07,
+      "loss": 0.1048,
+      "step": 472
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.304625080362939e-07,
+      "loss": 0.1365,
+      "step": 473
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.048367560708604e-07,
+      "loss": 0.2323,
+      "step": 474
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.79552098165296e-07,
+      "loss": 0.0435,
+      "step": 475
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.546094825649909e-07,
+      "loss": 0.0644,
+      "step": 476
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.300098446877925e-07,
+      "loss": 0.0884,
+      "step": 477
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.057541070889229e-07,
+      "loss": 0.1381,
+      "step": 478
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.818431794263837e-07,
+      "loss": 0.0472,
+      "step": 479
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.582779584268374e-07,
+      "loss": 0.0606,
+      "step": 480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.350593278519824e-07,
+      "loss": 0.0325,
+      "step": 481
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.121881584654056e-07,
+      "loss": 0.0391,
+      "step": 482
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.896653079999249e-07,
+      "loss": 0.0965,
+      "step": 483
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.67491621125429e-07,
+      "loss": 0.0288,
+      "step": 484
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.45667929417193e-07,
+      "loss": 0.0608,
+      "step": 485
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.241950513246931e-07,
+      "loss": 0.0619,
+      "step": 486
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.030737921409169e-07,
+      "loss": 0.2691,
+      "step": 487
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.823049439721562e-07,
+      "loss": 0.1071,
+      "step": 488
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.618892857083069e-07,
+      "loss": 0.1501,
+      "step": 489
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.418275829936537e-07,
+      "loss": 0.0807,
+      "step": 490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.221205881981594e-07,
+      "loss": 0.0666,
+      "step": 491
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.027690403892461e-07,
+      "loss": 0.0993,
+      "step": 492
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.837736653040825e-07,
+      "loss": 0.2467,
+      "step": 493
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6513517532236096e-07,
+      "loss": 0.0563,
+      "step": 494
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.468542694395861e-07,
+      "loss": 0.0792,
+      "step": 495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2893163324085886e-07,
+      "loss": 0.0648,
+      "step": 496
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.113679388751635e-07,
+      "loss": 0.3011,
+      "step": 497
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.941638450301644e-07,
+      "loss": 0.221,
+      "step": 498
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.773199969074959e-07,
+      "loss": 0.0961,
+      "step": 499
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.608370261985761e-07,
+      "loss": 0.0816,
+      "step": 500
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 541,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "total_flos": 1194534973440.0,
+  "train_batch_size": 10,
+  "trial_name": null,
+  "trial_params": null
+}

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39a1b7853796df569dea25e1cfc44a0eebd9c31bc3cf60c516910866a1ef6ae7
+size 6968

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/checkpoint-500/zero_to_fp32.py

@@ -0,0 +1,587 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel()
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file, tag=args.tag)

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/config.json

@@ -0,0 +1,70 @@
+{
+  "_name_or_path": "liuhaotian/llava-v1.6-mistral-7b",
+  "architectures": [
+    "LlavaMistralForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "freeze_mm_mlp_adapter": false,
+  "freeze_mm_vision_resampler": false,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "image_aspect_ratio": "anyres",
+  "image_crop_resolution": 224,
+  "image_grid_pinpoints": [
+    [
+      336,
+      672
+    ],
+    [
+      672,
+      336
+    ],
+    [
+      672,
+      672
+    ],
+    [
+      1008,
+      336
+    ],
+    [
+      336,
+      1008
+    ]
+  ],
+  "image_split_resolution": 224,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 32768,
+  "mm_hidden_size": 1024,
+  "mm_patch_merge_type": "spatial_unpad",
+  "mm_projector_lr": 2e-05,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "openai/clip-vit-large-patch14-336",
+  "mm_vision_tower_lr": 2e-06,
+  "model_type": "llava_mistral",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "rms_norm_eps": 1e-05,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "tokenizer_model_max_length": 4096,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.37.2",
+  "tune_mm_mlp_adapter": false,
+  "tune_mm_vision_resampler": false,
+  "unfreeze_mm_vision_tower": true,
+  "use_cache": true,
+  "use_mm_proj": true,
+  "vocab_size": 32000
+}

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/non_lora_trainables.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0102c05ff05f99863f06b141ed5812df27620d7c8dd7551f8bac60d6b2f9f0e
+size 41961648

CheckGuard Models/wholeimage/bank_no/llava-lora-mistral-r128a256-10BS-model/trainer_state.json

@@ -0,0 +1,3276 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 541,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.142857142857143e-07,
+      "loss": 0.6789,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4285714285714286e-06,
+      "loss": 0.8481,
+      "step": 2
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1428571428571427e-06,
+      "loss": 0.663,
+      "step": 3
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8571428571428573e-06,
+      "loss": 0.679,
+      "step": 4
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5714285714285718e-06,
+      "loss": 1.0166,
+      "step": 5
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2857142857142855e-06,
+      "loss": 0.4693,
+      "step": 6
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5e-06,
+      "loss": 0.4891,
+      "step": 7
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.7142857142857145e-06,
+      "loss": 0.5523,
+      "step": 8
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.4285714285714295e-06,
+      "loss": 0.2909,
+      "step": 9
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.1428571428571436e-06,
+      "loss": 0.2598,
+      "step": 10
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.857142857142858e-06,
+      "loss": 0.2532,
+      "step": 11
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.857142857142858e-06,
+      "loss": 0.4867,
+      "step": 12
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 0.4145,
+      "step": 13
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 0.3161,
+      "step": 14
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.285714285714288e-06,
+      "loss": 0.1836,
+      "step": 15
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-05,
+      "loss": 0.3355,
+      "step": 16
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0714285714285714e-05,
+      "loss": 0.2286,
+      "step": 17
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 0.3594,
+      "step": 18
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2142857142857142e-05,
+      "loss": 0.2981,
+      "step": 19
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2857142857142859e-05,
+      "loss": 0.3021,
+      "step": 20
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3571428571428574e-05,
+      "loss": 0.3866,
+      "step": 21
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4285714285714287e-05,
+      "loss": 0.2409,
+      "step": 22
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.1397,
+      "step": 23
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.5714285714285715e-05,
+      "loss": 0.1416,
+      "step": 24
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.642857142857143e-05,
+      "loss": 0.1838,
+      "step": 25
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7142857142857142e-05,
+      "loss": 0.1505,
+      "step": 26
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.785714285714286e-05,
+      "loss": 0.3278,
+      "step": 27
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8571428571428575e-05,
+      "loss": 0.2567,
+      "step": 28
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.928571428571429e-05,
+      "loss": 0.1218,
+      "step": 29
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2e-05,
+      "loss": 0.2288,
+      "step": 30
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9999812486015525e-05,
+      "loss": 0.1348,
+      "step": 31
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9999249951094388e-05,
+      "loss": 0.3734,
+      "step": 32
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999831241633323e-05,
+      "loss": 0.3169,
+      "step": 33
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9996999916892222e-05,
+      "loss": 0.1066,
+      "step": 34
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9995312501993765e-05,
+      "loss": 0.4434,
+      "step": 35
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9993250234920638e-05,
+      "loss": 0.198,
+      "step": 36
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9990813193013625e-05,
+      "loss": 0.115,
+      "step": 37
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9988001467668613e-05,
+      "loss": 0.2676,
+      "step": 38
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9984815164333163e-05,
+      "loss": 0.2201,
+      "step": 39
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9981254402502568e-05,
+      "loss": 0.1945,
+      "step": 40
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.997731931571535e-05,
+      "loss": 0.1391,
+      "step": 41
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9973010051548274e-05,
+      "loss": 0.2697,
+      "step": 42
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968326771610797e-05,
+      "loss": 0.1562,
+      "step": 43
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9963269651539018e-05,
+      "loss": 0.2204,
+      "step": 44
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9957838880989076e-05,
+      "loss": 0.2729,
+      "step": 45
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952034663630064e-05,
+      "loss": 0.441,
+      "step": 46
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952034663630064e-05,
+      "loss": 0.1401,
+      "step": 47
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9945857217136365e-05,
+      "loss": 0.3727,
+      "step": 48
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9939306773179498e-05,
+      "loss": 0.3269,
+      "step": 49
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9932383577419432e-05,
+      "loss": 0.0801,
+      "step": 50
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9925087889495374e-05,
+      "loss": 0.2772,
+      "step": 51
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9917419983016025e-05,
+      "loss": 0.2253,
+      "step": 52
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9909380145549325e-05,
+      "loss": 0.2318,
+      "step": 53
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9900968678611664e-05,
+      "loss": 0.1809,
+      "step": 54
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.989218589765658e-05,
+      "loss": 0.1155,
+      "step": 55
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9883032132062926e-05,
+      "loss": 0.2356,
+      "step": 56
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9873507725122505e-05,
+      "loss": 0.1194,
+      "step": 57
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9863613034027224e-05,
+      "loss": 0.3272,
+      "step": 58
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.985334842985567e-05,
+      "loss": 0.183,
+      "step": 59
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9842714297559212e-05,
+      "loss": 0.1217,
+      "step": 60
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9831711035947552e-05,
+      "loss": 0.1388,
+      "step": 61
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9820339057673773e-05,
+      "loss": 0.2112,
+      "step": 62
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9808598789218866e-05,
+      "loss": 0.0917,
+      "step": 63
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.979649067087574e-05,
+      "loss": 0.1585,
+      "step": 64
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9784015156732693e-05,
+      "loss": 0.1446,
+      "step": 65
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.97711727146564e-05,
+      "loss": 0.3511,
+      "step": 66
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9757963826274357e-05,
+      "loss": 0.1019,
+      "step": 67
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9744388986956824e-05,
+      "loss": 0.1165,
+      "step": 68
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.973044870579824e-05,
+      "loss": 0.2189,
+      "step": 69
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.971614350559814e-05,
+      "loss": 0.1254,
+      "step": 70
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.970147392284154e-05,
+      "loss": 0.0627,
+      "step": 71
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9686440507678827e-05,
+      "loss": 0.0952,
+      "step": 72
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.967104382390511e-05,
+      "loss": 0.1867,
+      "step": 73
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9655284448939094e-05,
+      "loss": 0.2003,
+      "step": 74
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9639162973801426e-05,
+      "loss": 0.1188,
+      "step": 75
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9622680003092503e-05,
+      "loss": 0.1111,
+      "step": 76
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.960583615496984e-05,
+      "loss": 0.1203,
+      "step": 77
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9588632061124837e-05,
+      "loss": 0.1599,
+      "step": 78
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9571068366759143e-05,
+      "loss": 0.209,
+      "step": 79
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9553145730560415e-05,
+      "loss": 0.2183,
+      "step": 80
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.953486482467764e-05,
+      "loss": 0.1351,
+      "step": 81
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.951622633469592e-05,
+      "loss": 0.128,
+      "step": 82
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9497230959610757e-05,
+      "loss": 0.2241,
+      "step": 83
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9477879411801843e-05,
+      "loss": 0.0991,
+      "step": 84
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9458172417006347e-05,
+      "loss": 0.1165,
+      "step": 85
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9438110714291697e-05,
+      "loss": 0.0792,
+      "step": 86
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9417695056027847e-05,
+      "loss": 0.121,
+      "step": 87
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9396926207859085e-05,
+      "loss": 0.2727,
+      "step": 88
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9375804948675308e-05,
+      "loss": 0.1947,
+      "step": 89
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.935433207058281e-05,
+      "loss": 0.2155,
+      "step": 90
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.933250837887457e-05,
+      "loss": 0.0525,
+      "step": 91
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9310334692000077e-05,
+      "loss": 0.2401,
+      "step": 92
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9287811841534598e-05,
+      "loss": 0.0743,
+      "step": 93
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9264940672148018e-05,
+      "loss": 0.1659,
+      "step": 94
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9241722041573166e-05,
+      "loss": 0.1184,
+      "step": 95
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9218156820573618e-05,
+      "loss": 0.1207,
+      "step": 96
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9194245892911077e-05,
+      "loss": 0.1292,
+      "step": 97
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.916999015531221e-05,
+      "loss": 0.2059,
+      "step": 98
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9145390517435013e-05,
+      "loss": 0.1682,
+      "step": 99
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9120447901834708e-05,
+      "loss": 0.1403,
+      "step": 100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9095163243929143e-05,
+      "loss": 0.1752,
+      "step": 101
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.906953749196371e-05,
+      "loss": 0.1616,
+      "step": 102
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9043571606975776e-05,
+      "loss": 0.1127,
+      "step": 103
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.901726656275866e-05,
+      "loss": 0.2236,
+      "step": 104
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8990623345825084e-05,
+      "loss": 0.2308,
+      "step": 105
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8963642955370203e-05,
+      "loss": 0.1739,
+      "step": 106
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8936326403234125e-05,
+      "loss": 0.1762,
+      "step": 107
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.890867471386395e-05,
+      "loss": 0.1457,
+      "step": 108
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.888068892427538e-05,
+      "loss": 0.2768,
+      "step": 109
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8852370084013783e-05,
+      "loss": 0.1389,
+      "step": 110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.882371925511488e-05,
+      "loss": 0.2747,
+      "step": 111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.879473751206489e-05,
+      "loss": 0.0542,
+      "step": 112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8765425941760237e-05,
+      "loss": 0.1414,
+      "step": 113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8735785643466786e-05,
+      "loss": 0.2482,
+      "step": 114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8705817728778626e-05,
+      "loss": 0.1602,
+      "step": 115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.867552332157637e-05,
+      "loss": 0.1342,
+      "step": 116
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8644903557985027e-05,
+      "loss": 0.077,
+      "step": 117
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8613959586331364e-05,
+      "loss": 0.0818,
+      "step": 118
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8582692567100866e-05,
+      "loss": 0.1443,
+      "step": 119
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855110367289421e-05,
+      "loss": 0.1148,
+      "step": 120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851919408838327e-05,
+      "loss": 0.1661,
+      "step": 121
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8486965010266726e-05,
+      "loss": 0.1676,
+      "step": 122
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845441764722514e-05,
+      "loss": 0.1288,
+      "step": 123
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.842155321987566e-05,
+      "loss": 0.0725,
+      "step": 124
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8388372960726228e-05,
+      "loss": 0.1258,
+      "step": 125
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8354878114129368e-05,
+      "loss": 0.068,
+      "step": 126
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8321069936235503e-05,
+      "loss": 0.1698,
+      "step": 127
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8286949694945864e-05,
+      "loss": 0.2038,
+      "step": 128
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8252518669864935e-05,
+      "loss": 0.0274,
+      "step": 129
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.821777815225245e-05,
+      "loss": 0.0564,
+      "step": 130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8182729444974993e-05,
+      "loss": 0.1182,
+      "step": 131
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8147373862457107e-05,
+      "loss": 0.3175,
+      "step": 132
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8111712730632024e-05,
+      "loss": 0.1017,
+      "step": 133
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.807574738689193e-05,
+      "loss": 0.3348,
+      "step": 134
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8039479180037803e-05,
+      "loss": 0.3129,
+      "step": 135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.800290947022884e-05,
+      "loss": 0.1095,
+      "step": 136
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7966039628931447e-05,
+      "loss": 0.1922,
+      "step": 137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7928871038867785e-05,
+      "loss": 0.1022,
+      "step": 138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.789140509396394e-05,
+      "loss": 0.2318,
+      "step": 139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7853643199297632e-05,
+      "loss": 0.2374,
+      "step": 140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7815586771045535e-05,
+      "loss": 0.1194,
+      "step": 141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.777723723643014e-05,
+      "loss": 0.1914,
+      "step": 142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.773859603366626e-05,
+      "loss": 0.0431,
+      "step": 143
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.769966461190707e-05,
+      "loss": 0.081,
+      "step": 144
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.766044443118978e-05,
+      "loss": 0.2162,
+      "step": 145
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.762093696238086e-05,
+      "loss": 0.1151,
+      "step": 146
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7581143687120877e-05,
+      "loss": 0.184,
+      "step": 147
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7541066097768965e-05,
+      "loss": 0.1963,
+      "step": 148
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.750070569734681e-05,
+      "loss": 0.1318,
+      "step": 149
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7460063999482314e-05,
+      "loss": 0.1163,
+      "step": 150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7419142528352815e-05,
+      "loss": 0.1013,
+      "step": 151
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.737794281862794e-05,
+      "loss": 0.0957,
+      "step": 152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7336466415412028e-05,
+      "loss": 0.2023,
+      "step": 153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.729471487418621e-05,
+      "loss": 0.1398,
+      "step": 154
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7252689760750053e-05,
+      "loss": 0.1238,
+      "step": 155
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.721039265116285e-05,
+      "loss": 0.2201,
+      "step": 156
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7167825131684516e-05,
+      "loss": 0.0698,
+      "step": 157
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7124988798716084e-05,
+      "loss": 0.0312,
+      "step": 158
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7081885258739846e-05,
+      "loss": 0.1443,
+      "step": 159
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7038516128259118e-05,
+      "loss": 0.1349,
+      "step": 160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6994883033737582e-05,
+      "loss": 0.0751,
+      "step": 161
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.695098761153832e-05,
+      "loss": 0.0543,
+      "step": 162
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6906831507862446e-05,
+      "loss": 0.0533,
+      "step": 163
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.686241637868734e-05,
+      "loss": 0.1328,
+      "step": 164
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6817743889704564e-05,
+      "loss": 0.3057,
+      "step": 165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6772815716257414e-05,
+      "loss": 0.1642,
+      "step": 166
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.672763354327804e-05,
+      "loss": 0.1479,
+      "step": 167
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6682199065224307e-05,
+      "loss": 0.1163,
+      "step": 168
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6636513986016215e-05,
+      "loss": 0.0395,
+      "step": 169
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6590580018972012e-05,
+      "loss": 0.0456,
+      "step": 170
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6544398886743934e-05,
+      "loss": 0.2018,
+      "step": 171
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.64979723212536e-05,
+      "loss": 0.1655,
+      "step": 172
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6451302063627067e-05,
+      "loss": 0.1805,
+      "step": 173
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6404389864129533e-05,
+      "loss": 0.2445,
+      "step": 174
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6357237482099682e-05,
+      "loss": 0.134,
+      "step": 175
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6309846685883726e-05,
+      "loss": 0.0976,
+      "step": 176
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6262219252769065e-05,
+      "loss": 0.0984,
+      "step": 177
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.621435696891765e-05,
+      "loss": 0.0495,
+      "step": 178
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6166261629298996e-05,
+      "loss": 0.1005,
+      "step": 179
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.6117935037622848e-05,
+      "loss": 0.1399,
+      "step": 180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.606937900627157e-05,
+      "loss": 0.2105,
+      "step": 181
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.6020595356232137e-05,
+      "loss": 0.142,
+      "step": 182
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5971585917027864e-05,
+      "loss": 0.0791,
+      "step": 183
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5922352526649803e-05,
+      "loss": 0.2,
+      "step": 184
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.587289703148779e-05,
+      "loss": 0.1317,
+      "step": 185
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5823221286261217e-05,
+      "loss": 0.1656,
+      "step": 186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5773327153949465e-05,
+      "loss": 0.3358,
+      "step": 187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.572321650572205e-05,
+      "loss": 0.2216,
+      "step": 188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.567289122086843e-05,
+      "loss": 0.0937,
+      "step": 189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5622353186727542e-05,
+      "loss": 0.0995,
+      "step": 190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.557160429861702e-05,
+      "loss": 0.2324,
+      "step": 191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5520646459762102e-05,
+      "loss": 0.2847,
+      "step": 192
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5469481581224274e-05,
+      "loss": 0.1242,
+      "step": 193
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5418111581829575e-05,
+      "loss": 0.1771,
+      "step": 194
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.536653838809667e-05,
+      "loss": 0.2115,
+      "step": 195
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.531476393416456e-05,
+      "loss": 0.074,
+      "step": 196
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5262790161720082e-05,
+      "loss": 0.0893,
+      "step": 197
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5210619019925066e-05,
+      "loss": 0.0644,
+      "step": 198
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5158252465343242e-05,
+      "loss": 0.2146,
+      "step": 199
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5105692461866874e-05,
+      "loss": 0.2579,
+      "step": 200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.50529409806431e-05,
+      "loss": 0.0806,
+      "step": 201
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.0806,
+      "step": 202
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4946871505372426e-05,
+      "loss": 0.132,
+      "step": 203
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4893557489227518e-05,
+      "loss": 0.1438,
+      "step": 204
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4840059950989992e-05,
+      "loss": 0.1703,
+      "step": 205
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.478638089696716e-05,
+      "loss": 0.0903,
+      "step": 206
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4732522340273686e-05,
+      "loss": 0.1515,
+      "step": 207
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.467848630075608e-05,
+      "loss": 0.2156,
+      "step": 208
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4624274804916958e-05,
+      "loss": 0.0783,
+      "step": 209
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.456988988583904e-05,
+      "loss": 0.1432,
+      "step": 210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4515333583108896e-05,
+      "loss": 0.1716,
+      "step": 211
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4460607942740468e-05,
+      "loss": 0.2328,
+      "step": 212
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4405715017098333e-05,
+      "loss": 0.1317,
+      "step": 213
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4350656864820733e-05,
+      "loss": 0.097,
+      "step": 214
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4295435550742372e-05,
+      "loss": 0.1547,
+      "step": 215
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4240053145816968e-05,
+      "loss": 0.0737,
+      "step": 216
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4184511727039612e-05,
+      "loss": 0.0926,
+      "step": 217
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4128813377368851e-05,
+      "loss": 0.0824,
+      "step": 218
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4072960185648576e-05,
+      "loss": 0.1236,
+      "step": 219
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.4016954246529697e-05,
+      "loss": 0.157,
+      "step": 220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.396079766039157e-05,
+      "loss": 0.1241,
+      "step": 221
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3904492533263243e-05,
+      "loss": 0.1243,
+      "step": 222
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3848040976744459e-05,
+      "loss": 0.1429,
+      "step": 223
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3791445107926478e-05,
+      "loss": 0.0321,
+      "step": 224
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3734707049312674e-05,
+      "loss": 0.0398,
+      "step": 225
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3677828928738934e-05,
+      "loss": 0.2625,
+      "step": 226
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3620812879293864e-05,
+      "loss": 0.0926,
+      "step": 227
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3563661039238785e-05,
+      "loss": 0.06,
+      "step": 228
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3506375551927546e-05,
+      "loss": 0.2397,
+      "step": 229
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3448958565726144e-05,
+      "loss": 0.157,
+      "step": 230
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3391412233932148e-05,
+      "loss": 0.1105,
+      "step": 231
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3333738714693958e-05,
+      "loss": 0.0877,
+      "step": 232
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3275940170929845e-05,
+      "loss": 0.1821,
+      "step": 233
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3218018770246858e-05,
+      "loss": 0.0166,
+      "step": 234
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.3159976684859528e-05,
+      "loss": 0.118,
+      "step": 235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.3101816091508389e-05,
+      "loss": 0.2289,
+      "step": 236
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.3043539171378362e-05,
+      "loss": 0.0518,
+      "step": 237
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2985148110016947e-05,
+      "loss": 0.1012,
+      "step": 238
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.292664509725226e-05,
+      "loss": 0.2009,
+      "step": 239
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2868032327110904e-05,
+      "loss": 0.252,
+      "step": 240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2809311997735697e-05,
+      "loss": 0.2044,
+      "step": 241
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2750486311303218e-05,
+      "loss": 0.1908,
+      "step": 242
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2691557473941246e-05,
+      "loss": 0.3064,
+      "step": 243
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2632527695645993e-05,
+      "loss": 0.091,
+      "step": 244
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.257339919019925e-05,
+      "loss": 0.0606,
+      "step": 245
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2514174175085346e-05,
+      "loss": 0.147,
+      "step": 246
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2454854871407993e-05,
+      "loss": 0.2029,
+      "step": 247
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.239544350380699e-05,
+      "loss": 0.0851,
+      "step": 248
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2335942300374788e-05,
+      "loss": 0.0904,
+      "step": 249
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2276353492572937e-05,
+      "loss": 0.0721,
+      "step": 250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2216679315148388e-05,
+      "loss": 0.1488,
+      "step": 251
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.2156922006049703e-05,
+      "loss": 0.1927,
+      "step": 252
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.2097083806343104e-05,
+      "loss": 0.029,
+      "step": 253
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.2037166960128443e-05,
+      "loss": 0.0301,
+      "step": 254
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1977173714455034e-05,
+      "loss": 0.1231,
+      "step": 255
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1917106319237386e-05,
+      "loss": 0.0348,
+      "step": 256
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1856967027170818e-05,
+      "loss": 0.0869,
+      "step": 257
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1796758093646989e-05,
+      "loss": 0.1164,
+      "step": 258
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1736481776669307e-05,
+      "loss": 0.0388,
+      "step": 259
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1676140336768236e-05,
+      "loss": 0.0433,
+      "step": 260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.161573603691655e-05,
+      "loss": 0.1996,
+      "step": 261
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1555271142444433e-05,
+      "loss": 0.2182,
+      "step": 262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1494747920954545e-05,
+      "loss": 0.0509,
+      "step": 263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1434168642236964e-05,
+      "loss": 0.1078,
+      "step": 264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1373535578184083e-05,
+      "loss": 0.0412,
+      "step": 265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1312851002705383e-05,
+      "loss": 0.2425,
+      "step": 266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.1252117191642175e-05,
+      "loss": 0.1119,
+      "step": 267
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1191336422682237e-05,
+      "loss": 0.0455,
+      "step": 268
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1130510975274408e-05,
+      "loss": 0.2613,
+      "step": 269
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1069643130543084e-05,
+      "loss": 0.0651,
+      "step": 270
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.1008735171202685e-05,
+      "loss": 0.1155,
+      "step": 271
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0947789381472035e-05,
+      "loss": 0.0661,
+      "step": 272
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0886808046988716e-05,
+      "loss": 0.0881,
+      "step": 273
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0825793454723325e-05,
+      "loss": 0.1123,
+      "step": 274
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0764747892893724e-05,
+      "loss": 0.14,
+      "step": 275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0703673650879219e-05,
+      "loss": 0.0889,
+      "step": 276
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0642573019134703e-05,
+      "loss": 0.1333,
+      "step": 277
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0581448289104759e-05,
+      "loss": 0.0608,
+      "step": 278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0520301753137725e-05,
+      "loss": 0.2882,
+      "step": 279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.045913570439972e-05,
+      "loss": 0.0661,
+      "step": 280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0397952436788643e-05,
+      "loss": 0.107,
+      "step": 281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0336754244848156e-05,
+      "loss": 0.0499,
+      "step": 282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0275543423681622e-05,
+      "loss": 0.237,
+      "step": 283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0214322268866033e-05,
+      "loss": 0.0301,
+      "step": 284
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0153093076365923e-05,
+      "loss": 0.0904,
+      "step": 285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0091858142447266e-05,
+      "loss": 0.0165,
+      "step": 286
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0030619763591348e-05,
+      "loss": 0.0791,
+      "step": 287
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.969380236408656e-06,
+      "loss": 0.1997,
+      "step": 288
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.908141857552737e-06,
+      "loss": 0.0155,
+      "step": 289
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.846906923634079e-06,
+      "loss": 0.0457,
+      "step": 290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.785677731133972e-06,
+      "loss": 0.0203,
+      "step": 291
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.724456576318383e-06,
+      "loss": 0.2384,
+      "step": 292
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.663245755151847e-06,
+      "loss": 0.1459,
+      "step": 293
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.602047563211359e-06,
+      "loss": 0.2249,
+      "step": 294
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.540864295600282e-06,
+      "loss": 0.037,
+      "step": 295
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.479698246862277e-06,
+      "loss": 0.145,
+      "step": 296
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.418551710895243e-06,
+      "loss": 0.1501,
+      "step": 297
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.3574269808653e-06,
+      "loss": 0.0727,
+      "step": 298
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.296326349120786e-06,
+      "loss": 0.0992,
+      "step": 299
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.23525210710628e-06,
+      "loss": 0.2516,
+      "step": 300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.174206545276678e-06,
+      "loss": 0.0628,
+      "step": 301
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.113191953011287e-06,
+      "loss": 0.132,
+      "step": 302
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.052210618527966e-06,
+      "loss": 0.0908,
+      "step": 303
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.991264828797319e-06,
+      "loss": 0.1432,
+      "step": 304
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.93035686945692e-06,
+      "loss": 0.0493,
+      "step": 305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.869489024725595e-06,
+      "loss": 0.0578,
+      "step": 306
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.808663577317765e-06,
+      "loss": 0.0909,
+      "step": 307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.747882808357828e-06,
+      "loss": 0.0646,
+      "step": 308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.687148997294622e-06,
+      "loss": 0.1308,
+      "step": 309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.626464421815919e-06,
+      "loss": 0.0729,
+      "step": 310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.565831357763039e-06,
+      "loss": 0.1871,
+      "step": 311
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.505252079045459e-06,
+      "loss": 0.1577,
+      "step": 312
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.444728857555572e-06,
+      "loss": 0.1844,
+      "step": 313
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.384263963083453e-06,
+      "loss": 0.1673,
+      "step": 314
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.323859663231768e-06,
+      "loss": 0.1898,
+      "step": 315
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.263518223330698e-06,
+      "loss": 0.1106,
+      "step": 316
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.203241906353014e-06,
+      "loss": 0.0476,
+      "step": 317
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.143032972829184e-06,
+      "loss": 0.1432,
+      "step": 318
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.082893680762619e-06,
+      "loss": 0.0249,
+      "step": 319
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.022826285544967e-06,
+      "loss": 0.0762,
+      "step": 320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.962833039871562e-06,
+      "loss": 0.1468,
+      "step": 321
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.902916193656898e-06,
+      "loss": 0.0272,
+      "step": 322
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.843077993950302e-06,
+      "loss": 0.0495,
+      "step": 323
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.783320684851613e-06,
+      "loss": 0.1958,
+      "step": 324
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.72364650742707e-06,
+      "loss": 0.0869,
+      "step": 325
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.664057699625215e-06,
+      "loss": 0.2957,
+      "step": 326
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.604556496193015e-06,
+      "loss": 0.0833,
+      "step": 327
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.545145128592009e-06,
+      "loss": 0.0978,
+      "step": 328
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.485825824914658e-06,
+      "loss": 0.1941,
+      "step": 329
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.426600809800753e-06,
+      "loss": 0.0384,
+      "step": 330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.367472304354011e-06,
+      "loss": 0.0872,
+      "step": 331
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.308442526058757e-06,
+      "loss": 0.1051,
+      "step": 332
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.249513688696786e-06,
+      "loss": 0.0918,
+      "step": 333
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.190688002264308e-06,
+      "loss": 0.2169,
+      "step": 334
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.131967672889101e-06,
+      "loss": 0.1647,
+      "step": 335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.073354902747742e-06,
+      "loss": 0.0585,
+      "step": 336
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.014851889983058e-06,
+      "loss": 0.1743,
+      "step": 337
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.956460828621641e-06,
+      "loss": 0.3001,
+      "step": 338
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.898183908491617e-06,
+      "loss": 0.0977,
+      "step": 339
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.840023315140476e-06,
+      "loss": 0.0549,
+      "step": 340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.781981229753145e-06,
+      "loss": 0.0738,
+      "step": 341
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.7240598290701585e-06,
+      "loss": 0.027,
+      "step": 342
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.666261285306048e-06,
+      "loss": 0.0647,
+      "step": 343
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.608587766067853e-06,
+      "loss": 0.0531,
+      "step": 344
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.551041434273862e-06,
+      "loss": 0.0582,
+      "step": 345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.4936244480724575e-06,
+      "loss": 0.2357,
+      "step": 346
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.4363389607612204e-06,
+      "loss": 0.0614,
+      "step": 347
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.379187120706138e-06,
+      "loss": 0.1516,
+      "step": 348
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.322171071261071e-06,
+      "loss": 0.2906,
+      "step": 349
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.265292950687329e-06,
+      "loss": 0.0402,
+      "step": 350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.208554892073528e-06,
+      "loss": 0.0895,
+      "step": 351
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.151959023255545e-06,
+      "loss": 0.109,
+      "step": 352
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.095507466736763e-06,
+      "loss": 0.1338,
+      "step": 353
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.039202339608432e-06,
+      "loss": 0.0541,
+      "step": 354
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.983045753470308e-06,
+      "loss": 0.0614,
+      "step": 355
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.927039814351426e-06,
+      "loss": 0.2844,
+      "step": 356
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.871186622631155e-06,
+      "loss": 0.1412,
+      "step": 357
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.815488272960388e-06,
+      "loss": 0.0575,
+      "step": 358
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.759946854183036e-06,
+      "loss": 0.1047,
+      "step": 359
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.704564449257635e-06,
+      "loss": 0.2065,
+      "step": 360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.649343135179271e-06,
+      "loss": 0.0995,
+      "step": 361
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.59428498290167e-06,
+      "loss": 0.1517,
+      "step": 362
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.539392057259536e-06,
+      "loss": 0.1122,
+      "step": 363
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.484666416891109e-06,
+      "loss": 0.0992,
+      "step": 364
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.430110114160965e-06,
+      "loss": 0.1303,
+      "step": 365
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.375725195083046e-06,
+      "loss": 0.1192,
+      "step": 366
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.321513699243924e-06,
+      "loss": 0.0991,
+      "step": 367
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.267477659726319e-06,
+      "loss": 0.077,
+      "step": 368
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.213619103032845e-06,
+      "loss": 0.1052,
+      "step": 369
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.159940049010015e-06,
+      "loss": 0.2359,
+      "step": 370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 5.106442510772489e-06,
+      "loss": 0.0501,
+      "step": 371
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 5.053128494627578e-06,
+      "loss": 0.0803,
+      "step": 372
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 5.000000000000003e-06,
+      "loss": 0.2073,
+      "step": 373
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.947059019356904e-06,
+      "loss": 0.0479,
+      "step": 374
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.89430753813313e-06,
+      "loss": 0.125,
+      "step": 375
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.8417475346567635e-06,
+      "loss": 0.0715,
+      "step": 376
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.78938098007494e-06,
+      "loss": 0.0242,
+      "step": 377
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.737209838279923e-06,
+      "loss": 0.1242,
+      "step": 378
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.685236065835443e-06,
+      "loss": 0.1771,
+      "step": 379
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.633461611903336e-06,
+      "loss": 0.1037,
+      "step": 380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.581888418170429e-06,
+      "loss": 0.0733,
+      "step": 381
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.530518418775734e-06,
+      "loss": 0.0565,
+      "step": 382
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.479353540237903e-06,
+      "loss": 0.1092,
+      "step": 383
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.4283957013829845e-06,
+      "loss": 0.0371,
+      "step": 384
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3776468132724605e-06,
+      "loss": 0.1105,
+      "step": 385
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.327108779131573e-06,
+      "loss": 0.1856,
+      "step": 386
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.276783494277954e-06,
+      "loss": 0.1237,
+      "step": 387
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.226672846050538e-06,
+      "loss": 0.2521,
+      "step": 388
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.176778713738787e-06,
+      "loss": 0.0565,
+      "step": 389
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.127102968512214e-06,
+      "loss": 0.0518,
+      "step": 390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.077647473350201e-06,
+      "loss": 0.0735,
+      "step": 391
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.028414082972141e-06,
+      "loss": 0.0786,
+      "step": 392
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.9794046437678705e-06,
+      "loss": 0.025,
+      "step": 393
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.930620993728434e-06,
+      "loss": 0.2235,
+      "step": 394
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.882064962377154e-06,
+      "loss": 0.1307,
+      "step": 395
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.83373837070101e-06,
+      "loss": 0.0224,
+      "step": 396
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7856430310823546e-06,
+      "loss": 0.1109,
+      "step": 397
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.737780747230941e-06,
+      "loss": 0.0624,
+      "step": 398
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.6901533141162804e-06,
+      "loss": 0.055,
+      "step": 399
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.6427625179003223e-06,
+      "loss": 0.2079,
+      "step": 400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.595610135870472e-06,
+      "loss": 0.2215,
+      "step": 401
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.548697936372937e-06,
+      "loss": 0.1016,
+      "step": 402
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5020276787464058e-06,
+      "loss": 0.1229,
+      "step": 403
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.455601113256073e-06,
+      "loss": 0.0759,
+      "step": 404
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.4094199810279926e-06,
+      "loss": 0.1667,
+      "step": 405
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.3634860139837877e-06,
+      "loss": 0.048,
+      "step": 406
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.317800934775696e-06,
+      "loss": 0.0543,
+      "step": 407
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2723664567219627e-06,
+      "loss": 0.1656,
+      "step": 408
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.2271842837425917e-06,
+      "loss": 0.0409,
+      "step": 409
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.1822561102954373e-06,
+      "loss": 0.1173,
+      "step": 410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.1375836213126653e-06,
+      "loss": 0.0964,
+      "step": 411
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0931684921375572e-06,
+      "loss": 0.0432,
+      "step": 412
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0490123884616795e-06,
+      "loss": 0.1451,
+      "step": 413
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.0051169662624224e-06,
+      "loss": 0.1226,
+      "step": 414
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.9614838717408866e-06,
+      "loss": 0.096,
+      "step": 415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.918114741260156e-06,
+      "loss": 0.1152,
+      "step": 416
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8750112012839215e-06,
+      "loss": 0.0575,
+      "step": 417
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8321748683154893e-06,
+      "loss": 0.097,
+      "step": 418
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7896073488371535e-06,
+      "loss": 0.0513,
+      "step": 419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7473102392499517e-06,
+      "loss": 0.0566,
+      "step": 420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7052851258137936e-06,
+      "loss": 0.0193,
+      "step": 421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.663533584587974e-06,
+      "loss": 0.1507,
+      "step": 422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.622057181372063e-06,
+      "loss": 0.0208,
+      "step": 423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.580857471647186e-06,
+      "loss": 0.0893,
+      "step": 424
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.539936000517689e-06,
+      "loss": 0.0988,
+      "step": 425
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4992943026531935e-06,
+      "loss": 0.0368,
+      "step": 426
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4589339022310386e-06,
+      "loss": 0.0911,
+      "step": 427
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4188563128791255e-06,
+      "loss": 0.1093,
+      "step": 428
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.379063037619146e-06,
+      "loss": 0.0717,
+      "step": 429
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.339555568810221e-06,
+      "loss": 0.1486,
+      "step": 430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.300335388092929e-06,
+      "loss": 0.1174,
+      "step": 431
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.261403966333742e-06,
+      "loss": 0.2022,
+      "step": 432
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.2227627635698624e-06,
+      "loss": 0.0376,
+      "step": 433
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1844132289544684e-06,
+      "loss": 0.3022,
+      "step": 434
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1463568007023706e-06,
+      "loss": 0.0121,
+      "step": 435
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.1085949060360654e-06,
+      "loss": 0.1441,
+      "step": 436
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0711289611322204e-06,
+      "loss": 0.0457,
+      "step": 437
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0339603710685574e-06,
+      "loss": 0.0324,
+      "step": 438
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9970905297711606e-06,
+      "loss": 0.045,
+      "step": 439
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9605208199621993e-06,
+      "loss": 0.0644,
+      "step": 440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.924252613108073e-06,
+      "loss": 0.0743,
+      "step": 441
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8882872693679787e-06,
+      "loss": 0.054,
+      "step": 442
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8526261375428955e-06,
+      "loss": 0.1679,
+      "step": 443
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8172705550250093e-06,
+      "loss": 0.0666,
+      "step": 444
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7822218477475496e-06,
+      "loss": 0.2,
+      "step": 445
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7474813301350668e-06,
+      "loss": 0.1191,
+      "step": 446
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7130503050541368e-06,
+      "loss": 0.1166,
+      "step": 447
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6789300637645e-06,
+      "loss": 0.0089,
+      "step": 448
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6451218858706374e-06,
+      "loss": 0.0848,
+      "step": 449
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6116270392737753e-06,
+      "loss": 0.1263,
+      "step": 450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.578446780124344e-06,
+      "loss": 0.1338,
+      "step": 451
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5455823527748626e-06,
+      "loss": 0.0566,
+      "step": 452
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5130349897332764e-06,
+      "loss": 0.0618,
+      "step": 453
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4808059116167306e-06,
+      "loss": 0.0259,
+      "step": 454
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4488963271057943e-06,
+      "loss": 0.1682,
+      "step": 455
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4173074328991376e-06,
+      "loss": 0.0967,
+      "step": 456
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3860404136686411e-06,
+      "loss": 0.0799,
+      "step": 457
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.355096442014977e-06,
+      "loss": 0.1426,
+      "step": 458
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.3244766784236307e-06,
+      "loss": 0.1401,
+      "step": 459
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.294182271221377e-06,
+      "loss": 0.0526,
+      "step": 460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2642143565332154e-06,
+      "loss": 0.1516,
+      "step": 461
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2345740582397647e-06,
+      "loss": 0.0326,
+      "step": 462
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.2052624879351105e-06,
+      "loss": 0.0517,
+      "step": 463
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.176280744885121e-06,
+      "loss": 0.094,
+      "step": 464
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1476299159862204e-06,
+      "loss": 0.0684,
+      "step": 465
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1193110757246251e-06,
+      "loss": 0.0845,
+      "step": 466
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.09132528613605e-06,
+      "loss": 0.1105,
+      "step": 467
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0636735967658785e-06,
+      "loss": 0.0947,
+      "step": 468
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0363570446297999e-06,
+      "loss": 0.0685,
+      "step": 469
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0093766541749206e-06,
+      "loss": 0.0902,
+      "step": 470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.827334372413444e-07,
+      "loss": 0.0257,
+      "step": 471
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.564283930242258e-07,
+      "loss": 0.1048,
+      "step": 472
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.304625080362939e-07,
+      "loss": 0.1365,
+      "step": 473
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.048367560708604e-07,
+      "loss": 0.2323,
+      "step": 474
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.79552098165296e-07,
+      "loss": 0.0435,
+      "step": 475
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.546094825649909e-07,
+      "loss": 0.0644,
+      "step": 476
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.300098446877925e-07,
+      "loss": 0.0884,
+      "step": 477
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.057541070889229e-07,
+      "loss": 0.1381,
+      "step": 478
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.818431794263837e-07,
+      "loss": 0.0472,
+      "step": 479
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.582779584268374e-07,
+      "loss": 0.0606,
+      "step": 480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.350593278519824e-07,
+      "loss": 0.0325,
+      "step": 481
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 7.121881584654056e-07,
+      "loss": 0.0391,
+      "step": 482
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.896653079999249e-07,
+      "loss": 0.0965,
+      "step": 483
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.67491621125429e-07,
+      "loss": 0.0288,
+      "step": 484
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.45667929417193e-07,
+      "loss": 0.0608,
+      "step": 485
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.241950513246931e-07,
+      "loss": 0.0619,
+      "step": 486
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.030737921409169e-07,
+      "loss": 0.2691,
+      "step": 487
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.823049439721562e-07,
+      "loss": 0.1071,
+      "step": 488
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.618892857083069e-07,
+      "loss": 0.1501,
+      "step": 489
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.418275829936537e-07,
+      "loss": 0.0807,
+      "step": 490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.221205881981594e-07,
+      "loss": 0.0666,
+      "step": 491
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.027690403892461e-07,
+      "loss": 0.0993,
+      "step": 492
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.837736653040825e-07,
+      "loss": 0.2467,
+      "step": 493
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6513517532236096e-07,
+      "loss": 0.0563,
+      "step": 494
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.468542694395861e-07,
+      "loss": 0.0792,
+      "step": 495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2893163324085886e-07,
+      "loss": 0.0648,
+      "step": 496
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.113679388751635e-07,
+      "loss": 0.3011,
+      "step": 497
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.941638450301644e-07,
+      "loss": 0.221,
+      "step": 498
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.773199969074959e-07,
+      "loss": 0.0961,
+      "step": 499
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.608370261985761e-07,
+      "loss": 0.0816,
+      "step": 500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4471555106090573e-07,
+      "loss": 0.0565,
+      "step": 501
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2895617609489337e-07,
+      "loss": 0.0703,
+      "step": 502
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.135594923211771e-07,
+      "loss": 0.0622,
+      "step": 503
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9852607715846194e-07,
+      "loss": 0.1138,
+      "step": 504
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.838564944018618e-07,
+      "loss": 0.2741,
+      "step": 505
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6955129420176193e-07,
+      "loss": 0.06,
+      "step": 506
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.556110130431788e-07,
+      "loss": 0.0322,
+      "step": 507
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.420361737256438e-07,
+      "loss": 0.1867,
+      "step": 508
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2882728534360131e-07,
+      "loss": 0.2815,
+      "step": 509
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.159848432673084e-07,
+      "loss": 0.133,
+      "step": 510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.035093291242607e-07,
+      "loss": 0.134,
+      "step": 511
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.914012107811336e-07,
+      "loss": 0.1156,
+      "step": 512
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7966094232622856e-07,
+      "loss": 0.3407,
+      "step": 513
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6828896405244988e-07,
+      "loss": 0.0641,
+      "step": 514
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.572857024407881e-07,
+      "loss": 0.0459,
+      "step": 515
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.466515701443294e-07,
+      "loss": 0.1403,
+      "step": 516
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3638696597277678e-07,
+      "loss": 0.0836,
+      "step": 517
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.264922748774955e-07,
+      "loss": 0.1507,
+      "step": 518
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1696786793707782e-07,
+      "loss": 0.1091,
+      "step": 519
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0781410234342093e-07,
+      "loss": 0.0432,
+      "step": 520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.90313213883376e-08,
+      "loss": 0.0166,
+      "step": 521
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.061985445067756e-08,
+      "loss": 0.1675,
+      "step": 522
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.258001698397744e-08,
+      "loss": 0.0462,
+      "step": 523
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.491211050462798e-08,
+      "loss": 0.0219,
+      "step": 524
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.761642258056977e-08,
+      "loss": 0.1261,
+      "step": 525
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.069322682050516e-08,
+      "loss": 0.1249,
+      "step": 526
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.414278286363761e-08,
+      "loss": 0.0674,
+      "step": 527
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.796533636993728e-08,
+      "loss": 0.0171,
+      "step": 528
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.216111901092501e-08,
+      "loss": 0.0653,
+      "step": 529
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.6730348460986e-08,
+      "loss": 0.0292,
+      "step": 530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.167322838920406e-08,
+      "loss": 0.1442,
+      "step": 531
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6989948451726643e-08,
+      "loss": 0.0773,
+      "step": 532
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2680684284650532e-08,
+      "loss": 0.0428,
+      "step": 533
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8745597497433765e-08,
+      "loss": 0.2392,
+      "step": 534
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.518483566683826e-08,
+      "loss": 0.1413,
+      "step": 535
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1998532331389812e-08,
+      "loss": 0.0554,
+      "step": 536
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.186806986376528e-09,
+      "loss": 0.1174,
+      "step": 537
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.749765079363535e-09,
+      "loss": 0.048,
+      "step": 538
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.687498006236135e-09,
+      "loss": 0.1818,
+      "step": 539
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0000831077803273e-09,
+      "loss": 0.083,
+      "step": 540
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6875836667729073e-09,
+      "loss": 0.0186,
+      "step": 541
+    },
+    {
+      "epoch": 1.0,
+      "step": 541,
+      "total_flos": 1291092221952.0,
+      "train_loss": 0.14259492732281495,
+      "train_runtime": 4024.6248,
+      "train_samples_per_second": 1.342,
+      "train_steps_per_second": 0.134
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 541,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "total_flos": 1291092221952.0,
+  "train_batch_size": 10,
+  "trial_name": null,
+  "trial_params": null
+}

CheckGuard Models/wholeimage/check_no/finetune_lora_llava_mistral.sh

@@ -0,0 +1,43 @@
+#!/bin/bash
+# Use first parameter as GPU IDs, default to "0,1,2,3" if not provided
+GPU_IDS=${1:-0,1,2,3}
+
+
+CUDA_VISIBLE_DEVICES=0,1,2,3 deepspeed --include localhost:"$GPU_IDS" --master_port 29604\
+    llava/train/train_mem.py \
+    --lora_enable True --lora_r 128 --lora_alpha 256 --mm_projector_lr 2e-5 \
+    --deepspeed ./scripts/zero3.json \
+    --model_name_or_path liuhaotian/llava-v1.6-mistral-7b \
+    --version mistral_instruct \
+    --data_path /home/larry5/project/LLaVA-1.6-ft/data/peft/check_no/check_no_dataset.json \
+    --image_folder /home/larry5/project/LLaVA-1.6-ft/data/data/ \
+    --vision_tower openai/clip-vit-large-patch14-336 \
+    --mm_projector_type mlp2x_gelu \
+    --mm_vision_select_layer -2 \
+    --mm_use_im_start_end False \
+    --mm_use_im_patch_token False \
+    --mm_patch_merge_type spatial_unpad \
+    --image_aspect_ratio anyres \
+    --group_by_modality_length False \
+    --bf16 False \
+    --fp16 True \
+    --output_dir /home/larry5/project/LLaVA-1.6-ft/scripts_peft/mistral/lora/llava-lora-mistral-r128a256/wholeimage/check_no/llava-lora-mistral-r128a256-10BS-model \
+    --num_train_epochs 1 \
+    --per_device_train_batch_size 10 \
+    --per_device_eval_batch_size 1 \
+    --gradient_accumulation_steps 1 \
+    --evaluation_strategy "no" \
+    --save_strategy "steps" \
+    --save_steps 500 \
+    --save_total_limit 5 \
+    --learning_rate 2e-5 \
+    --weight_decay 0. \
+    --warmup_ratio 0.05 \
+    --lr_scheduler_type "cosine" \
+    --logging_steps 1 \
+    --tf32 True \
+    --model_max_length 4096 \
+    --gradient_checkpointing True \
+    --dataloader_num_workers 4 \
+    --lazy_preprocess True \
+    --report_to wandb \